WO2021108969A1 - 属性信息的预测方法、编码器、解码器、及存储介质 - Google Patents
属性信息的预测方法、编码器、解码器、及存储介质 Download PDFInfo
- Publication number
- WO2021108969A1 WO2021108969A1 PCT/CN2019/122499 CN2019122499W WO2021108969A1 WO 2021108969 A1 WO2021108969 A1 WO 2021108969A1 CN 2019122499 W CN2019122499 W CN 2019122499W WO 2021108969 A1 WO2021108969 A1 WO 2021108969A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- point
- predicted
- adjacent
- morton code
- attribute
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/001—Model-based coding, e.g. wire frame
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
- G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/004—Predictors, e.g. intraframe, interframe coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/13—Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/46—Embedding additional information in the video signal during the compression process
- H04N19/463—Embedding additional information in the video signal during the compression process by compressing encoding parameters before transmission
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/597—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding specially adapted for multi-view video sequence encoding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
- H04N19/88—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression involving rearrangement of data among different coding units, e.g. shuffling, interleaving, scrambling or permutation of pixel data or permutation of transform coefficient data among different blocks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/91—Entropy coding, e.g. variable length coding [VLC] or arithmetic coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/96—Tree coding, e.g. quad-tree coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
Definitions
- the embodiments of the present application relate to coding and decoding technologies in the communication field, and in particular, to a method for predicting attribute information, an encoder, a decoder, and a storage medium.
- the input point cloud can be divided into geometric information and attribute information corresponding to each point.
- the geometric information of the point cloud and the location of each point cloud The corresponding attribute information is coded separately.
- the coordinate information of the point can be used to obtain the Morton code corresponding to each point in the point cloud, and then based on the Morton code, the attribute information is predicted by the backward difference method.
- Morton codes have periodic jump points, it is spatially inaccurate to only use adjacent points of Morton codes to predict the current point, which affects the accuracy of prediction and leads to prediction residuals. Large defects, which in turn reduce coding efficiency.
- the embodiments of the present application provide a prediction method, an encoder, a decoder, and a storage medium for attribute information, which can effectively improve the accuracy of prediction, greatly reduce prediction residuals, and thereby improve coding efficiency.
- an embodiment of the present application provides a method for predicting attribute information, which is applied to an encoder, and the method includes:
- the target Morton code corresponding to the point to be predicted is determined according to a preset neighbor information table; wherein, the preset neighbor information table is used to determine the neighboring position of the point to be predicted Determine the coordinate difference with the preset reference point;
- an embodiment of the present application also provides a method for predicting attribute information, which is applied to a decoder, and the method includes:
- the target Morton code corresponding to the point to be predicted is determined according to a preset neighbor information table; wherein, the preset neighbor information table is used to determine the neighboring position of the point to be predicted Determine the coordinate difference with the preset reference point;
- an embodiment of the present application also provides an encoder.
- the encoder includes: a first determining part, a first judging part, and a first predicting part,
- the first determining part is configured to determine the current Morton code of the point to be predicted in the point cloud to be coded; and based on the current Morton code, determine the target corresponding to the point to be predicted according to a preset neighbor information table Morton code; wherein the preset neighbor information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the first judgment part is configured to judge whether there are adjacent points of the point to be predicted in the point cloud to be coded according to the target Morton code
- the first prediction part is configured to predict the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the adjacent point if the adjacent point exists in the point cloud to be coded.
- an embodiment of the present application also provides a decoder, which includes: a second determining part, a second determining part, and a second predicting part,
- the second determining part is configured to determine the current Morton code of the point to be predicted in the point cloud to be decoded; and based on the current Morton code, determine the target corresponding to the point to be predicted according to a preset neighbor information table Morton code; wherein the preset neighbor information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the second determining part is configured to determine whether there are adjacent points of the point to be predicted in the point cloud to be decoded according to the target Morton code;
- the second prediction part is configured to predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point if the adjacent point exists in the point cloud to be decoded.
- an embodiment of the present application also provides an encoder.
- the encoder includes a first processor and a first memory storing executable instructions of the first processor. When a processor is executed, the method for predicting attribute information as described above is realized.
- an embodiment of the present application also provides a decoder.
- the decoder includes a second processor and a second memory storing executable instructions of the second processor. When the second processor executes, the method for predicting attribute information as described above is realized.
- an embodiment of the present application also provides a computer-readable storage medium with a program stored thereon and applied to an encoder.
- the program is executed by a processor, the method for predicting attribute information as described above is implemented. .
- the embodiments of the present application also provide a computer-readable storage medium with a program stored thereon and applied to a decoder.
- the program is executed by a processor, the method for predicting attribute information as described above is implemented. .
- the embodiment of the application provides a method for predicting attribute information, an encoder, a decoder, and a storage medium.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be encoded; based on the current Morton code, according to the preset
- the adjacent information table determines the target Morton code corresponding to the point to be predicted; the preset adjacent information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code It is judged whether there are adjacent points of the point to be predicted in the point cloud to be coded; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the decoder determines the current Morton code of the point to be predicted in the point cloud to be decoded; based on the current Morton code, determines the target Morton code corresponding to the point to be predicted according to the preset neighbor information table; judges the target Morton code to be decoded according to the target Morton code Whether there are adjacent points of the point to be predicted in the point cloud; if there are adjacent points in the point cloud to be decoded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- Figure 1 is a schematic diagram of the coding process of PCEM
- FIG. 2 is a schematic diagram of the decoding process of PCEM
- Figure 3 shows the spatial coding of each pixel of an 8 ⁇ 8 image
- Figure 4 shows the spatial coding of each pixel of a 4 ⁇ 4 image
- Figure 5 is the first schematic diagram of the application of the prediction method
- Figure 6 is a second schematic diagram of the application of the prediction method
- FIG. 7 is a schematic diagram 1 of the implementation process of the method for predicting attribute information
- Figure 8 is a schematic diagram of adjacent positions
- Fig. 9 is a second schematic diagram of the realization process of the attribute information prediction method.
- Figure 10 is the third schematic diagram of the implementation process of the attribute information prediction method
- Figure 11 is a fourth schematic diagram of the implementation process of the attribute information prediction method.
- Fig. 12 is a fifth schematic diagram of the realization process of the attribute information prediction method
- Figure 13 is a schematic diagram of a prediction framework for attribute information
- Fig. 14 is a flowchart of a prediction algorithm for attribute information
- Figure 15 is a sixth schematic diagram of the implementation process of the attribute information prediction method
- Fig. 16 is a seventh schematic diagram of the realization process of the method for predicting attribute information
- Figure 17 is the eighth schematic diagram of the implementation process of the attribute information prediction method.
- Figure 18 is a schematic diagram of the composition of the encoder
- Figure 19 is a second schematic diagram of the composition of the encoder
- Figure 20 is a schematic diagram of the structure of the decoder
- Figure 21 is a second schematic diagram of the structure of the decoder.
- the input point cloud can be divided into geometric information and attribute information corresponding to each point.
- the geometric information of the point cloud and the attribute information corresponding to each point cloud are separately coded.
- Figure 1 is a schematic diagram of the encoding process of PCEM.
- the geometric information is coordinated so that all the point clouds are contained in a bounding box (bounding box), and then quantization is performed.
- Quantization mainly plays the role of scaling. Due to quantization rounding, the geometric information of a part of the point cloud is the same. Therefore, it is determined whether to remove duplicate points based on parameters.
- the process of quantization and removal of duplicate points is also called voxelization. process.
- the geometric information is reconstructed to guide the attribute coding.
- attribute coding is mainly carried out for color information.
- the color information (that is, the attribute information) is converted from the RGB (Red-Green-Blue) color space to the YUV (Luminance-Chrominance) color space.
- RGB Red-Green-Blue
- YUV Luminance-Chrominance
- Fig. 2 is a schematic diagram of the decoding process of PCEM.
- the geometric bit stream and the attribute bit stream in the binary code stream are decoded independently.
- decoding the geometric bitstream obtain the geometric information of the point cloud through entropy decoding-octree reconstruction-inverse coordinate quantization-inverse coordinate translation;
- decoding the attribute bitstream through entropy decoding-inverse quantization-attribute Reconstruction-inverse space transformation obtains the attribute information of the point cloud, and restores the three-dimensional image model of the point cloud data to be encoded based on the geometric information and the attribute information.
- the coordinate information of the point can be used to obtain the Morton code corresponding to each point in the point cloud, and then based on the Morton code, the attribute information is processed by the backward difference method. prediction.
- Morton coding is also called z-order code, because its coding order is based on spatial z order. The specific method of calculating Morton code is described as follows. For each component of the three-dimensional coordinate represented by a d-bit binary number, the representation of the three coordinate components is realized by the following:
- Morton code M is to start from the highest position of x, y, z, and alternately arrange x l , y l , z l to the lowest position in turn.
- the calculation formula of M is as follows:
- Figure 3 shows the spatial encoding of each pixel of an 8 ⁇ 8 image.
- the position coordinates of x and y values from 0-7 are encoded with one-dimensional binary numbers.
- Interleave binary coordinate values to obtain a binary z-value map.
- Connect the z-shape along the numerical direction to produce a recursive z-shaped curve.
- the z value is placed in the order of connection.
- the above figure is iteratively generated in the z direction, from 00-11 (one z in the entire figure), then from 0000-1111 (one z for each point of the previous z), and then from 000000-111111 (the previous Put one z) for each point of z, increase by two bits each time, and increase recursively.
- Figure 4 shows the spatial coding of each pixel of a 4 ⁇ 4 image.
- the coordinates x and y are binarized first, and then the highest position of x and y At the beginning, let them alternately arrange the values of each bit from high to low.
- connect the Morton code values from small to large you can see the zigzag connection sequence, and every four points can be regarded as a small block, except for the last two digits, the first four digits are the same.
- the same Morton code of the first four bits can determine a parent block.
- the Morton code of the parent block can be obtained by shifting the Morton code of these four points to the right by 2 digits.
- every four blocks of the same level are continuously combined, that is, the Morton code value is shifted by two bits to the right, and their parent block can be obtained.
- every eight small blocks are combined, that is, the Morton code value is shifted three places to the right, and their parent blocks can be obtained.
- the points adjacent to the Morton code can be used to predict the current point.
- the specific process is as follows:
- Step3. Repeat Step2 until the complete frame point cloud is traversed.
- Morton codes have periodic transition points, that is to say, the corresponding spatial positions of two adjacent Morton codes are not necessarily adjacent. Therefore, only the adjacent points of Morton codes are used to Prediction of the current point is spatially inaccurate, that is, in PCEM, only the previous Morton code is used to predict the attribute information of the current point, there will be errors, which will affect the accuracy of the prediction.
- this application proposes to use the current Morton code of the point to be predicted to quickly find the true neighboring point with stronger correlation with the point to be predicted, and use the attribute reconstruction value of the neighboring point to determine the attributes of the predicted point.
- Information is predicted to improve the performance of the encoder. That is to say, in this application, when the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the neighbors around the point to be predicted. The target Morton code corresponding to the position.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the attribute of the point to be predicted
- the predicted value is precisely because the correlation between the point to be predicted and the neighboring points is stronger, so the prediction of the attribute information of the point to be predicted through the neighboring points can effectively improve the accuracy of the prediction and greatly reduce the prediction. Residual, thereby improving coding efficiency.
- FIG. 5 is a schematic diagram of the application of the prediction method, as shown in Figure 5.
- the attribute prediction method of the present application can be applied to the attribute prediction position in the PCEM point cloud encoder framework.
- Fig. 6 is a second schematic diagram of the application of the prediction method.
- the attribute prediction method of the present application can be applied to the attribute reconstruction position in the PCEM point cloud decoder framework.
- the prediction method of the attribute information acts on the encoding end and the decoding end simultaneously, and the process, framework and algorithm of the prediction method of the attribute information are exactly the same on the encoding end and the decoding end.
- FIG. 7 is a schematic diagram of the implementation process of the method for predicting attribute information. As shown in FIG. 7, the encoder predicts the prediction point
- the method of attribute information may include the following steps:
- Step 101 Determine the current Morton code of the point to be predicted in the point cloud to be coded.
- the encoder may first determine the current Morton code of the point to be predicted in the point cloud to be coded.
- the geometric information of the point cloud to be coded and the attribute information corresponding to the point cloud to be coded are separately coded. Among them, after the geometric information encoding is completed, the geometric information is reconstructed, and the encoding of the attribute information will depend on the reconstructed geometric information.
- the coding of attribute information is mainly for the coding of color information.
- the color information is converted from the RGB color space to the YUV color space.
- use the reconstructed geometric information to recolor the point cloud, so that the uncoded attribute information corresponds to the reconstructed geometric information.
- the attribute information is predicted.
- the point cloud to be coded is reordered based on the Morton code, and a point cloud sequence that can be used to predict the attribute information of the point cloud to be coded is generated.
- the encoder can use the set information of each point in the point cloud to be coded to determine the Morton code of each point. After obtaining the Morton code of each point, the encoder can sort all the points in the point cloud to be coded in ascending order, sort all the Morton codes of all points, and generate the corresponding point cloud to be coded. Point cloud sequence collection. Among them, all Morton codes of all points in the point cloud to be coded are stored in the point cloud sequence set.
- the encoder when the encoder encodes the attribute information of the point cloud to be coded, it can predict all the points of the point cloud to be coded in the order of Morton code from small to large. .
- the point to be predicted can be any point in the point cloud to be coded
- the current Morton code is the Morton code of the point to be predicted
- the attribute information of the point to be predicted is predicted
- other points corresponding to other Morton codes smaller than the current Morton code have completed the attribute information prediction, and the corresponding attribute reconstruction information is generated.
- the encoder when the encoder predicts the attribute information of the point to be predicted, it needs to determine the Morton code of the point to be predicted first to obtain the current Morton code.
- Step 102 Based on the current Morton code, determine the target Morton code corresponding to the point to be predicted according to the preset neighbor information table; wherein, the preset neighbor information table is used for the relationship between the neighboring position of the point to be predicted and the preset reference point. The coordinate difference between is determined.
- the encoder may determine the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table.
- the target Morton code can represent the Morton code corresponding to the neighboring position around the point to be predicted.
- the preset neighbor information table can represent the coordinate difference between the neighboring position around the point to be predicted and the preset reference point.
- the adjacent position of the point to be predicted may include coplanar adjacent points that are coplanar adjacent to the point to be predicted, and may also include collinear adjacent points that are collinearly adjacent to the point to be predicted.
- the preset reference point when predicting the attribute information of the point to be predicted, if the point to be predicted is taken as the spatial center of the point cloud to be coded, then the preset reference point may be the point to be predicted as The coordinate origin of the center of the space. Specifically, based on the same preset reference point, the difference between the distance between the preset reference point and the point to be predicted and the distance between the preset reference point and the adjacent position of the point to be predicted is calculated to obtain The coordinate difference between the adjacent position and the point to be predicted.
- the preset reference point may be the point with the smallest Morton code in the point cloud to be coded.
- Figure 8 is a schematic diagram of adjacent positions. As shown in Figure 8, the preset reference point is O, the point to be predicted is A, and B, C and D that are coplanar adjacent to A are coplanar adjacent points. E, F, and G that are adjacent to each other are collinear adjacent points.
- the encoder performs octree segmentation. Therefore, for the point to be predicted, there are 26 adjacent positions around it according to the unit distance. However, because the encoder treats When encoding the attribute information of the code point cloud, the attribute information is predicted in the order of Morton code from small to large. Therefore, the encoder only needs to consider the points whose Morton code is smaller than the current Morton code. For example, Just consider B, C, D, E, F, and G in Figure 2 above.
- the encoder before the encoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table, it needs to first establish the preset phase corresponding to the point to be predicted. Neighbor information table. In other words, the preset neighbor information table is established based on the points to be predicted.
- the encoder when the encoder establishes the preset neighbor information table corresponding to the point to be predicted, it can operate on the current Morton code (cur_pos) of the point to be predicted. Specifically, the encoder You can first locate the Morton code (base_pos) of the preset reference point, and then calculate the coordinate difference between the preset reference point and the adjacent position of the point to be predicted based on the Morton code of the preset reference point, and then continue Morton coding is performed on these coordinate differences, so that a preset neighbor information table corresponding to the points to be predicted can be established.
- the adjacent positions of the points to be predicted may include coplanar adjacent points that are coplanar adjacent to the point to be predicted, and collinear adjacent points that are coplanar and adjacent to the point to be predicted. Therefore, when the encoder establishes the preset neighbor information table corresponding to the point to be predicted, it can respectively calculate the first coordinate difference between the preset reference point and the coplanar adjacent points, and calculate the preset reference point and the collinear phase. For the second coordinate difference between adjacent points, Morton encoding can be performed on the first coordinate difference to obtain the coplanar Morton code corresponding to the coplanar adjacent points.
- Morton encoding can be performed on the second coordinate difference to obtain The collinear Morton code corresponding to the collinear adjacent points.
- the encoder can establish a preset neighbor information table corresponding to the point to be predicted based on the coplanar Morton code and the collinear Morton code.
- the encoder may determine the to-be-predicted based on the preset neighbor information table corresponding to the current Morton code.
- the encoder when the encoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table, since the preset neighbor information table stores the target Morton code, Morton code after Morton encoding is performed on the coordinate difference between the preset reference point and the adjacent position. Therefore, the encoder can extract the Morton of the coordinate difference corresponding to the adjacent position from the preset adjacent information table. Code, the Morton code of the coordinate difference can be further used to determine the target Morton code.
- the encoder is performing the target Morton code
- the encoder is performing target Morton code
- the adjacent type first, where the adjacent type may include coplanar adjacent and collinear adjacent, and then based on the determined adjacent type, the common adjacent information table can be further extracted. Planar Morton code or collinear Morton code, and determine the target Morton code based on coplanar Morton code or collinear Morton code.
- the distance between the coplanar adjacent point and it is 1, then the distance between the coplanar adjacent point and the point is Based on the principle that the closer the distance, the stronger the correlation, the adjacent types of coplanar adjacent can be selected first, and then the coplanar Morton code can be extracted from the preset adjacent information table, so as to use the coplanar Morton code to determine the target model. Pause.
- the target Morton code can represent the Morton code corresponding to the neighboring position around the point to be predicted. Therefore, the encoder can use the target Morton code to treat the neighboring neighboring points around the prediction point. Click to find and confirm.
- Step 103 Determine whether there are adjacent points of the point to be predicted in the point cloud to be coded according to the target Morton code.
- the encoder after the encoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table, it can determine the point cloud to be encoded according to the target Morton code Whether there are adjacent points to the point to be predicted. In other words, the encoder can use Morton codes corresponding to adjacent positions around the point to be predicted to further determine whether there are points adjacent to the point to be predicted in the coded point cloud.
- the encoder determines the target Morton code, it cannot directly determine the neighboring points of the point to be predicted based on the target Morton code. This is because the point cloud is parallel in space. It is not evenly distributed, and the adjacent points of the points to be predicted do not necessarily have adjacent points. That is to say, the target Morton code determined according to the adjacent positions of the points to be predicted may not really be in the point cloud to be coded. Therefore, after determining the target Morton code, the encoder needs to determine whether there are adjacent points of the point to be predicted in the point cloud to be coded according to the target Morton code.
- the encoder after obtaining the Morton code of each point, the encoder will sort all the Morton codes of all points in ascending order to generate the points to be coded.
- the point cloud sequence set corresponding to the cloud Therefore, after the encoder determines the target Morton code, it can use the point cloud sequence set corresponding to the point cloud to be encoded, combined with the target Morton code, to further determine whether there are adjacent points to be predicted Click to judge.
- the encoder may first obtain the point cloud of the point cloud to be coded when judging whether there are adjacent points of the point to be predicted in the coded point cloud according to the point cloud sequence set and the target Morton code. Sequence collection, and then use the target Morton code to perform query processing in the point cloud sequence collection, and the query result can be obtained. If the query result is that the target Morton code exists in the point cloud sequence collection, the encoder can determine the point cloud to be coded There are adjacent points corresponding to the points to be predicted. That is to say, in this application, after determining the target Morton code, the encoder directly searches for the existence of the target Morton code in the point cloud sequence set. If it exists, it can be considered that the point to be predicted has adjacent points in the space. ; Correspondingly, if it does not exist, it can be considered that the point to be predicted does not have an adjacent point in the space.
- the encoder when it performs query processing in the point cloud sequence set based on the target Morton code, it can directly traverse each query point cloud sequence set according to the target Morton code. Morton code to obtain the query result; according to the target Morton code, the query processing of the point cloud sequence set can also be performed according to the preset query algorithm to obtain the query result.
- the encoder when the encoder performs query processing in the point cloud sequence collection based on the target Morton code, it can either use the traversal method used for the point cloud sequence collection, or it can be other pairs of points. Cloud sequence collection for fast search algorithm.
- Step 104 If there are adjacent points in the point cloud to be coded, predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point.
- the encoder judges whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are adjacent points of the point to be predicted in the point cloud to be coded, then the encoder The information can be directly reconstructed based on the attribute information of the adjacent points of the point to be predicted, and the attribute information of the point to be predicted can be predicted to obtain the attribute prediction value.
- the encoder may first obtain the neighboring points when predicting the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the neighboring points. Then calculate the mean value of the reconstructed attribute value to obtain the mean value corresponding to the reconstructed attribute value, and then the mean value can be determined as the predictive value of the attribute.
- the adjacent points of the points to be predicted may be coplanar adjacent points that are coplanar and adjacent to the point to be predicted, or they may be collinearly adjacent to the point to be predicted.
- the encoder determines that the adjacent points of the points to be predicted are coplanar adjacent points or collinear adjacent points, then the encoder can obtain the co-linear adjacent points when obtaining the attribute reconstruction values of the adjacent points. The reconstructed value of the first attribute corresponding to the neighboring points of the plane, or obtaining the reconstructed value of the second attribute corresponding to the collinear neighbors.
- the encoder is performing the reconstruction of the attribute value.
- the first average value corresponding to at least one first reconstruction value can be calculated, or the second average value corresponding to at least one second reconstruction value can be calculated.
- the encoder can calculate the first average value or the second average value. The value is determined as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the encoder determines that the adjacent points of the point to be predicted are coplanar adjacent points and collinear adjacent points, then when the encoder obtains the attribute reconstruction values of the adjacent points, The first attribute reconstruction value corresponding to the coplanar adjacent points can be obtained, and at the same time, the second attribute reconstruction value corresponding to the collinear adjacent points can be obtained.
- the encoder is performing the reconstruction of the attribute value.
- a weighted average calculation can be performed on the reconstructed value of the first attribute and the reconstructed value of the second attribute according to the preset weight, so that the weighted average can be obtained.
- the encoder can determine the weighted average value as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the encoder when the encoder performs a weighted average calculation on the reconstruction value of the first attribute and the reconstruction value of the second attribute by using the preset weight, the reconstruction value of the first attribute and the reconstruction value of the second attribute can be calculated separately.
- the reconstruction values are assigned different weight values. Specifically, based on the principle that the closer the distance, the stronger the correlation, the weight value corresponding to the first attribute reconstruction value should be greater than the weight value corresponding to the second attribute reconstruction value.
- the point to be predicted has multiple adjacent points such as coplanar adjacent points and collinear adjacent points, for example, there are 6 adjacent points to predict the attribute information of the point to be predicted, including 3 coplanar adjacent points and 3 collinear adjacent points.
- the encoder can perform a distance-based weighted average operation on the attribute reconstruction values of all adjacent points, so as to realize the prediction of the attribute information of the point to be predicted, and obtain the attribute prediction value of the point to be predicted.
- the encoder can use formula (2) to predict the attribute information of the points to be predicted:
- d ij is the distance between the adjacent point and the point to be predicted
- f(d ij ) is a function related to d ij , that is, according to the distance between each adjacent point and the point to be predicted
- the encoder when the encoder predicts the attribute information of the point to be predicted, it can simultaneously use the first attribute reconstruction value and the collinear phase of the coplanar adjacent points of the point to be predicted.
- the second attribute reconstruction value of the neighboring point determines the attribute prediction value of the point to be predicted.
- the attribute predicted value of the point to be predicted Among them, based on the principle that the closer the distance, the stronger the correlation, the closer coplanar neighboring points can be preferentially selected to predict the attribute information of the points to be predicted.
- the encoder may first search for coplanar adjacent points with a distance of 1, and if it exists, use the average value of the attribute reconstruction values of these coplanar adjacent points as the attribute prediction value of the point to be predicted ; If there is no coplanar adjacent point of the point to be predicted in the point cloud to be coded, then continue to search for a distance of Collinear adjacent points, and then use the average value of the attribute reconstruction values of these collinear adjacent points as the attribute prediction value of the point to be predicted.
- Figure 9 is a schematic diagram of the second implementation process of the attribute information prediction method.
- the encoder determines whether there are adjacent points in the coded point cloud according to the target Morton code, that is, step After 103, the method for the encoder to predict the attribute information of the point to be predicted may further include the following steps:
- Step 105 If there are no adjacent points in the coded point cloud, use the previous Morton code corresponding to the current Morton code to predict the attribute prediction value of the point to be predicted.
- the encoder after the encoder judges whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are no adjacent points of the point to be predicted in the point cloud to be coded, then the code Therefore, the encoder needs to use the previous Morton code corresponding to the current Morton code to predict the attribute information of the point to be predicted, so as to obtain the attribute information of the point to be predicted. The predicted value of the attribute of the point.
- the encoder when the encoder uses the previous Morton code corresponding to the current Morton code to predict the attribute prediction value of the point to be predicted, it may first determine the previous Morton code from the point cloud to be coded For the corresponding reconstructed point, the attribute reconstruction value of the reconstructed point can then be obtained, and the attribute reconstruction value of the reconstructed point can be determined as the attribute prediction value of the point to be predicted.
- the attribute reconstruction value of the point is used as the attribute prediction value of the point to be predicted to complete the prediction of the attribute information of the point to be predicted.
- the encoder can first search for the current Morton code from the point cloud sequence set of the point cloud to be coded The previous Morton code. Among them, all Morton codes of all points in the point cloud to be coded are stored in the point cloud sequence set in ascending order.
- Figure 10 is a schematic diagram of the third implementation process of the attribute information prediction method.
- the encoder determines whether there are adjacent points in the coded point cloud according to the target Morton code, that is, step After 103 and before predicting the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the neighboring points, that is, before step 104, the method for the encoder to predict the attribute information of the point to be predicted may further include the following steps:
- Step 106 Determine the reconstructed point corresponding to the previous Morton code from the point cloud to be coded.
- Step 107 Determine the reconstructed point as the neighboring point of the point to be predicted.
- the encoder after the encoder judges whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are no adjacent points of the point to be predicted in the point cloud to be coded, that is, In other words, there are no coplanar adjacent points and collinear adjacent points around the point to be predicted, so the encoder can first determine the reconstructed point corresponding to the previous Morton code from the point cloud to be coded.
- the encoder can determine the neighboring point of the point to be predicted by the reconstructed point. That is to say, in this application, based on the target Morton code, if there is no adjacent point of the point to be predicted in the point cloud to be coded, then directly reconstruct the corresponding Morton code of the current Morton code
- the points are regarded as adjacent points of the points to be predicted, and then the method of step 104 is followed to complete the prediction of the attribute information of the points to be predicted.
- step 103 that is, after the encoder determines whether there are adjacent points in the point cloud to be encoded according to the target Morton code, if there are points to be predicted in the point cloud to be encoded
- the encoder can directly predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point; if there is no adjacent point of the point to be predicted in the point cloud to be coded, the encoder can start from the point to be predicted.
- the adjacent points of the points can continue to predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent points.
- the encoder can first compare the current Morton in the point cloud sequence set.
- the reconstructed point corresponding to the previous Morton code of the frame is regarded as the neighboring point of the point to be predicted, and then the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the neighboring point.
- the encoder can calculate the prediction residual of the point to be predicted according to the attribute prediction value, and then The prediction residual can be coded to generate a bitstream.
- the encoder predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the information of the point to be predicted can be used.
- the current Morton code determines the target Morton code corresponding to the adjacent position around the point to be predicted. If there is an adjacent point to the point to be predicted in the adjacent position in the point cloud, the codec can directly use the phase.
- the attribute reconstruction value prediction of the neighboring points obtains the attribute prediction value of the point to be predicted. It is precisely because the correlation between the point to be predicted and the neighboring point is stronger, so the prediction of the attribute information of the point to be predicted can be achieved through the neighboring points.
- this application directly obtains the Morton code of the neighboring point according to the current Morton code of the point to be predicted, and when it is determined that the neighboring point does exist after the query, the reconstructed attribute value of the neighboring point is used to treat the predicted point It uses the spatial correlation of the point cloud to a greater extent than the related technical solutions of PCEM.
- PSNR Peak Signal to Noise Ratio
- the encoder can also apply the method from step 101 to step 107 to the intra-layer prediction of the point cloud to be coded.
- the method for predicting attribute information proposed in the present application can make full use of the spatial correlation of point clouds, so that the prediction residuals are smaller, and it is more suitable for subsequent quantization and entropy coding. Under the premise of PSNR performance, higher gain is achieved.
- Table 1 is the performance of the attribute information prediction method proposed by this application. It can be seen from Table 1 that compared with the related technical solutions of PCEM, for different targets, under the same objective quality, the attribute information proposed by this application is used. In the prediction method, the code rate is reduced, and the performance of the encoder is improved.
- the embodiment of the application provides a method for predicting attribute information.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be coded; based on the current Morton code, the corresponding point to be predicted is determined according to the preset neighbor information table.
- the target Morton code wherein, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code, it is judged whether there is a point cloud to be coded.
- the adjacent points of the predicted point; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- FIG. 11 is a fourth schematic diagram of the implementation process of the method for predicting attribute information.
- the adjacent position of the point to be predicted may include the same as the point to be predicted.
- Step 201 Position a preset reference point.
- Step 202 Calculate the first coordinate difference between the preset reference point and the coplanar adjacent points, and calculate the second coordinate difference between the preset reference point and the coplanar adjacent points;
- Step 203 Perform Morton coding on the first coordinate difference to obtain the coplanar Morton code corresponding to the coplanar adjacent points, and perform Morton coding on the second coordinate difference to obtain the collinear Morton code corresponding to the collinear adjacent points ;
- Step 204 Based on the coplanar Morton code and the collinear Morton code, establish a preset neighbor information table corresponding to the point to be predicted.
- the encoder when the encoder establishes the preset neighbor information table, it needs to locate the preset reference point first, where the preset reference point represents the space centered on the point to be predicted in the point cloud to be coded. Origin coordinates.
- the preset reference point may be the point with the smallest Morton code in the point cloud to be coded.
- the coordinates of the point to be predicted are (1, 1, 2), and with the point to be predicted as the center of the space, the encoder may position the preset reference point coordinates as (0, 0, 0) .
- the coordinates of the neighboring positions around the point to be predicted may be determined based on the preset reference point, for example, a certain neighboring position is determined
- the coordinates of is (1, 1, 1).
- the coordinate difference between the adjacent positions of the preset reference point and the point to be predicted may specifically include the first coordinate difference between the preset reference point and the coplanar adjacent points, It may also include the second coordinate difference between the preset reference point and the collinear adjacent point.
- the encoder uses the preset reference point to calculate the coordinate difference between the preset reference point and the adjacent position, it can perform Morton encoding on the coordinate difference. , Get the corresponding Morton code.
- the encoder after the encoder calculates the first coordinate difference between the preset reference point and the coplanar adjacent point, it can perform Morton encoding on the first coordinate difference to obtain the corresponding Morton code, namely coplanar Morton code; correspondingly, the encoder can perform Morton coding on the second coordinate difference after calculating the second coordinate difference between the preset reference point and the collinear adjacent point, so as to obtain the corresponding Morton code, that is, collinear Morton code.
- the encoder after the encoder completes Morton encoding of the coordinate difference between the preset reference point and the adjacent position, it can establish preset adjacent information based on the obtained Morton code table. Specifically, the encoder can establish a preset neighboring information table by using the correspondence between neighboring positions and Morton codes. In other words, the preset neighbor information table can be used to determine Morton codes at neighboring positions.
- the encoder can establish a preset neighbor information table corresponding to the point to be predicted based on the coplanar Morton code and the collinear Morton code. Specifically, the encoder can Preset the adjacent information table for the coplanar adjacent points of the predicted points and their corresponding coplanar Morton code storage values. At the same time, the collinear adjacent points of the points to be predicted and their corresponding collinear Morton codes can be stored The value presets the adjacent information table, thereby completing the establishment of the preset adjacent information table.
- FIG. 12 is a schematic diagram five of the implementation process of the attribute information prediction method.
- the encoder determines the point to be predicted based on the current Morton code according to the preset neighbor information table
- the corresponding target Morton code method may include the following steps:
- Step 102a Read the coplanar Morton code and the collinear Morton code corresponding to the point to be predicted from the preset neighbor information table.
- Step 102b Perform a summation operation on the coplanar Morton code and the Morton code of the preset reference point to obtain a first sum value, and perform a summation operation on the collinear Morton code and the Morton code of the preset reference point to obtain The second sum value.
- Step 102c Determine the first sum value and the second sum value as the target Morton code.
- the encoder when the encoder uses the current Morton code and the preset neighbor information table corresponding to the point to be predicted to determine the target Morton code, it can first read the total value from the preset neighbor information table.
- Surface Morton code and collinear Morton code that is to say, the encoder can first obtain the Morton code corresponding to the coplanar adjacent point of the point to be predicted from the preset adjacent information table, that is, the coplanar Morton code.
- the Morton code corresponding to the collinear adjacent point of the point to be predicted can be obtained from the preset adjacent information table, that is, the collinear Morton code.
- the coplanar Morton code and the collinear Morton code can be further used to determine the target Morton code.
- the encoder can perform a summation operation on the coplanar Morton code and the Morton code of the preset reference point to obtain the first sum value.
- the coplanar Morton code can represent the distance difference between the preset reference point and the coplanar adjacent point
- the encoder adds the coplanar Morton code to the Morton code of the preset reference point.
- the Morton code of the coplanar adjacent points can be directly obtained, that is, the first sum value can be used as the Morton code of the coplanar adjacent points of the point to be predicted, so the first sum value can be determined as the target Morton code .
- the encoder can perform a summation operation on the collinear Morton code and the Morton code of the preset reference point to obtain the second sum value.
- the collinear Morton code can represent the distance difference between the preset reference point and the collinear adjacent points
- the encoder adds the collinear Morton code to the Morton code of the preset reference point.
- the Morton code of the collinear adjacent points can be directly obtained, that is, the second sum value can be used as the Morton code of the collinear adjacent point of the point to be predicted, so the second sum value can be determined as the target Morton code .
- the code of the point to be predicted may be determined first.
- the adjacent type is selected to determine whether the adjacent type is a coplanar type or a collinear type. In other words, the encoder can choose to acquire only the Morton codes of the coplanar adjacent points of the points to be predicted, or choose to acquire only the Morton codes of the collinear adjacent points of the points to be predicted.
- the method for the encoder to determine the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table may include the following steps:
- Step 102d If the adjacent type is a coplanar type, read the coplanar Morton code corresponding to the point to be predicted from the preset adjacent information table.
- Step 102e Perform a summation operation on the coplanar Morton code and the Morton code of the preset reference point to obtain a first sum value.
- Step 102f Determine the first sum value as the target Morton code.
- the encoder if the encoder first selects the adjacent type as coplanar adjacent, then the encoder only needs to read the coplanar Morton code corresponding to the point to be predicted from the preset adjacent information table, and then correct The coplanar Morton code and the Morton code of the preset reference point are summed to obtain the first sum value, and finally the first sum value can be determined as the target Morton code.
- the encoder chooses to obtain only the Morton code of the coplanar adjacent points of the point to be predicted, the encoder is using the current Morton code and the preset phase corresponding to the point to be predicted.
- the neighbor information table determines the target Morton code
- the coplanar Morton code corresponding to the coplanar neighboring point of the point to be predicted can be first obtained from the preset neighbor information table. Then, the sum of the coplanar Morton code and the Morton code of the preset reference point can be further performed.
- the coplanar Morton code can represent the distance difference between the preset reference point and the coplanar adjacent point, therefore, After the encoder adds the coplanar Morton code and the Morton code of the preset reference point, the first sum value can be determined as the target Morton code.
- the encoder can preferentially select the adjacent type as coplanar adjacent, so that the prediction residual can be reduced, Reduce the output binary code stream as much as possible to improve coding efficiency.
- the method for the encoder to determine the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table may further include the following steps:
- Step 102g If the adjacent type is a collinear type, read the collinear Morton code corresponding to the point to be predicted from the preset adjacent information table.
- Step 102h Perform a summation operation on the collinear Morton code and the Morton code of the preset reference point to obtain a second sum value.
- Step 102i Determine the second sum value as the target Morton code.
- the encoder if the encoder first selects the adjacent type as collinear adjacent, then the encoder only needs to read the collinear Morton code corresponding to the point to be predicted from the preset adjacent information table, and then correct The collinear Morton code and the Morton code of the preset reference point are summed to obtain the second sum value, and finally the second sum value can be determined as the target Morton code.
- the encoder chooses to obtain only the Morton codes of the collinear adjacent points of the points to be predicted, the encoder is using the current Morton code and the preset phase corresponding to the points to be predicted.
- the collinear Morton code corresponding to the collinear adjacent point of the point to be predicted may be first obtained from the preset adjacent information table. Then the collinear Morton code and the Morton code of the preset reference point can be further summed.
- the collinear Morton code can represent the distance difference between the preset reference point and the collinear adjacent points, therefore, After the encoder adds the collinear Morton code and the Morton code of the preset reference point, the second sum value can be determined as the target Morton code.
- the encoder can preferentially select the adjacent type as coplanar adjacent, but if the point cloud to be coded does not exist in the point cloud to be coded, the similarity is stronger. If the prediction points are coplanar adjacent points, then the encoder can select the adjacent type as collinear adjacent points, so that the prediction residuals can be reduced as much as possible to improve the coding efficiency.
- the embodiment of the application provides a method for predicting attribute information.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be coded; based on the current Morton code, the corresponding point to be predicted is determined according to the preset neighbor information table.
- the target Morton code wherein, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code, it is judged whether there is a point cloud to be coded.
- the adjacent points of the predicted point; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the neighboring point is stronger, so the prediction of the attribute information of the point to be predicted through the neighboring point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- the adjacent points may specifically include coplanar adjacent points that are coplanar and adjacent to the point to be predicted.
- the adjacent points may also include collinear adjacent points that are collinearly adjacent to the point to be predicted. Therefore, the method for the encoder to predict the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the adjacent point may include the following steps:
- Step 301 If the neighboring point is a coplanar neighboring point that is coplanar neighboring with the point to be predicted, obtain the first attribute reconstruction value corresponding to the coplanar neighboring point.
- Step 302 Calculate the first average value of the first attribute reconstruction value, and determine the first average value as the attribute predicted value.
- the encoder after the encoder judges whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are adjacent points of the point to be predicted in the point cloud to be coded, and there are The adjacent points of the point to be predicted are only coplanar adjacent points that are coplanar adjacent to the point to be predicted, then the encoder can obtain the first attribute corresponding to the coplanar adjacent point when obtaining the attribute reconstruction value of the adjacent point Rebuild value.
- the encoder can calculate at least one first reconstruction when calculating the mean value of the attribute reconstruction value. The value corresponds to the first average value, and then the encoder can determine the first average value as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the encoder can use the following formula (4) to calculate the average value of the attribute reconstruction values of all adjacent points to obtain the point to be predicted. Predicted value of attribute of predicted point:
- a 1j is the reconstructed value of the attribute of the coplanar adjacent point.
- the method for the encoder to predict the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the neighboring point may include the following steps:
- Step 303 If the adjacent points are collinear adjacent points that are collinear adjacent to the point to be predicted, obtain the second attribute reconstruction value corresponding to the collinear adjacent point.
- Step 304 Calculate a second average value of the second attribute reconstruction value, and determine the second average value as the attribute predicted value.
- the encoder after the encoder judges whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are adjacent points of the point to be predicted in the point cloud to be coded, and there are The adjacent points of the point to be predicted are only the collinear adjacent points that are collinearly adjacent to the point to be predicted, then the encoder can obtain the second attribute corresponding to the collinear adjacent point when obtaining the attribute reconstruction value of the adjacent point Rebuild value.
- the encoder can calculate at least one second reconstruction when calculating the mean value of the attribute reconstruction value. The second average value corresponding to the value, and then the encoder can determine the second average value as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the encoder can use the following formula (5) to calculate the average value of the attribute reconstruction values of all adjacent points to obtain the point to be predicted. Predicted value of attribute of predicted point:
- I the predicted value of the attribute of the current point
- a 2j is the reconstructed value of the attribute of the collinear adjacent point.
- the method for the encoder to predict the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the neighboring point may include the following steps:
- Step 305 If the adjacent points are coplanar adjacent points that are coplanar adjacent to the point to be predicted, and collinear adjacent points that are collinearly adjacent to the point to be predicted, obtain the first coplanar adjacent point corresponding to the The reconstructed value of the attribute, and the reconstructed value of the second attribute corresponding to the collinear adjacent point.
- Step 306 Perform a weighted average calculation on the reconstructed value of the first attribute and the reconstructed value of the second attribute according to the preset weight to obtain a weighted average.
- Step 307 Determine the weighted average value as the attribute prediction value.
- the encoder determines whether there are adjacent points of the point to be predicted in the coded point cloud according to the target Morton code, if there are adjacent points of the point to be predicted in the point cloud to be coded, and the point to be predicted is
- the adjacent points of a point include both coplanar adjacent points and collinear adjacent points, so when the encoder obtains the attribute reconstruction value of the adjacent point, it can obtain the first attribute reconstruction value corresponding to the coplanar adjacent point, At the same time, the second attribute reconstruction value corresponding to the collinear adjacent points is obtained.
- the encoder since the point to be predicted has at least one coplanar adjacent coplanar adjacent point, and at least one collinear adjacent collinear adjacent point, the encoder is in the When calculating the average of the reconstructed value, the first attribute reconstruction value and the second attribute reconstruction value may be calculated on a weighted average according to the preset weight, so that the weighted average can be obtained.
- the encoder can assign different weight values to the first attribute reconstruction value and the second attribute reconstruction value based on the principle that the closer the distance, the stronger the correlation. Specifically, the weight value corresponding to the first attribute reconstruction value should be greater than the weight value corresponding to the second attribute reconstruction value.
- the encoder when the encoder predicts the attribute information of the point to be predicted, it can use both the first attribute reconstruction value of the coplanar adjacent point of the point to be predicted and the first attribute reconstruction value of the coplanar adjacent point of the point to be predicted.
- the second attribute reconstruction value determines the attribute prediction value of the point to be predicted. You can also choose to use the first attribute reconstruction value of the coplanar adjacent points of the point to be predicted, or the second attribute reconstruction value of the collinear adjacent points to determine the attribute prediction value of the point to be predicted. Attribute predicted value.
- the encoder can first search for coplanar adjacent points with a distance of 1, and if they exist, use the average value of the attribute reconstruction values of these coplanar adjacent points as the attribute prediction value of the point to be predicted; if the point cloud to be coded is There is no coplanar adjacent point of the point to be predicted, then the search distance is Collinear adjacent points, and then use the average of the attribute reconstruction values of these collinear adjacent points as the attribute predicted value of the point to be predicted.
- the embodiment of the application provides a method for predicting attribute information.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be coded; based on the current Morton code, the corresponding point to be predicted is determined according to the preset neighbor information table.
- the target Morton code wherein, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code, it is judged whether there is a point cloud to be coded.
- the adjacent points of the predicted point; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the neighboring point is stronger, so the prediction of the attribute information of the point to be predicted through the neighboring point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- FIG. 13 is a schematic diagram of the prediction framework of attribute information.
- the method for the encoder to predict the attribute information by using the current Morton code may include the following steps:
- Step 401 Calculate the current Morton code of the point to be predicted.
- the encoder can first determine the Morton code of the point to be predicted in the point cloud to be coded, that is, the current Morton code.
- Step 402 Select the adjacent type of the adjacent point.
- the encoder can select the adjacent type before determining the target Morton code of the adjacent point.
- the adjacent type may include the coplanar type and the collinear type, and then based on the adjacent type, the selection of the coplanar Morton code or the collinear Morton code is further obtained.
- the encoder cannot determine whether there are adjacent points of the point to be predicted in the point cloud to be encoded. Therefore, the adjacent points here are assumed to be predicted. There are points around the neighboring positions.
- Step 403 Determine the target Morton code of the neighboring point based on the neighboring type.
- the adjacent type selected by the encoder is coplanar adjacent, then the coplanar Morton code corresponding to the coplanar adjacent point can be obtained from the pre-established preset adjacent information table and used as the target Morton code ; If the adjacent type selected by the encoder is collinear adjacent, then the collinear Morton code corresponding to the collinear adjacent point can be obtained from the pre-established preset adjacent information table, and it is used as the target Morton code.
- Step 404 Determine whether there are adjacent points, if there are, then execute step 405; if not, then execute step 406.
- the encoder can use the target Morton code to determine whether there are adjacent points in the point cloud to be encoded. Specifically, the encoder can perform query processing in the point cloud sequence set of the point cloud to be encoded, that is, directly search in the point cloud sequence set Whether there is a target Morton code, if it exists, it can be considered that the point to be predicted has an adjacent point in the space; correspondingly, if it does not exist, it can be considered that the point to be predicted does not have an adjacent point in the space.
- Step 405 Predict the attribute information of the point to be predicted by using the reconstructed value of the attribute of the neighboring point.
- the encoder directly obtains the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the adjacent point of the point to be predicted.
- the encoder may perform average calculation on the attribute reconstruction values of adjacent points, and determine the average value of the attribute reconstruction values as the attribute prediction value.
- Step 406 Use the previous Morton code of the current Morton code to predict the attribute information of the point to be predicted.
- the encoder needs to use the previous Morton code corresponding to the current Morton code to obtain the attribute prediction value of the point to be predicted.
- the encoder can determine the reconstructed point corresponding to the previous Morton code, and then can obtain the attribute reconstruction value of the reconstructed point, and determine the attribute reconstruction value of the reconstructed point as the attribute prediction value of the point to be predicted.
- Step 407 Calculate the prediction residual corresponding to the point to be predicted.
- Step 408 Generate a code stream.
- the encoder After the encoder predicts and obtains the attribute prediction value of the point to be predicted, it can use the attribute prediction value to calculate the prediction residual, and then encode the prediction residual to obtain and output a binary code stream.
- FIG. 14 another exemplary implementation of predicting attribute information is shown in FIG. 14 as a flowchart of the attribute information prediction algorithm.
- the specific implementation process is It can include the following steps:
- Step 501 Determine the current Morton code Cur_pos of the point to be predicted.
- the encoder can first determine the Morton code of the point to be predicted in the point cloud to be coded, that is, the current Morton code Cur_pos.
- Step 502 Locate the Morton code base_pos of the preset reference point.
- the encoder operates on Cur_pos and locates to the Morton code base_pos of the reference point. Wherein, the encoder can pre-calculate the coordinate difference between the reference point base_pos and the adjacent point of the point to be predicted, perform Morton coding on these differences, and establish a preset table.
- Step 503 Calculate and obtain the target Morton code neighbor_pos of the neighboring points based on base_pos.
- Step 504 Based on neighbor_pos, search for the existence of the neighboring point Neighbor, if it does, then directly execute step 506; if it does not exist, execute step 505 first, and then continue to execute step 506.
- the encoder can use the target Morton code to determine whether there are adjacent points in the point cloud to be encoded. Specifically, the encoder can perform query processing in the point cloud sequence set of the point cloud to be encoded, that is, directly search in the point cloud sequence set Whether there is a target Morton code, if it exists, it can be considered that the point to be predicted has an adjacent point in the space; correspondingly, if it does not exist, it can be considered that the point to be predicted does not have an adjacent point in the space.
- Step 505 Use the point corresponding to the last Morton code lastMortonCode of the current Morton code as the adjacent point of the point to be predicted.
- the encoder needs to use the last Morton code corresponding to the current Morton code to predict the attribute information of the point to be predicted.
- the previous Morton code The frame code is the Morton code of the adjacent point of the point to be predicted, that is, the point corresponding to the last Morton code lastMortonCode of the current Morton code is regarded as the adjacent point of the point to be predicted.
- Neighbor lastMortonCode.
- Step 506 Determine the attribute prediction value Pred of the point to be predicted by using the reconstructed value neighbor[i] of the neighboring point.
- Step 507 Calculate the prediction residual Residual of the point to be predicted based on Pred.
- the encoder After the encoder predicts and obtains the attribute prediction value of the point to be predicted, it can use the attribute prediction value to calculate the prediction residual, and then encode the prediction residual to generate and output the attribute code stream.
- the encoder can first compare the current Morton in the point cloud sequence set.
- the reconstructed point corresponding to the previous Morton code of the frame is regarded as the neighboring point of the point to be predicted, and then the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the neighboring point.
- the embodiment of the application provides a method for predicting attribute information.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be coded; based on the current Morton code, the corresponding point to be predicted is determined according to the preset neighbor information table.
- the target Morton code wherein, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code, it is judged whether there is a point cloud to be coded.
- the adjacent points of the predicted point; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- FIG. 15 is a schematic diagram of the implementation process of the method for predicting attribute information. As shown in FIG. 15, the decoder predicts the point to be predicted.
- the method of attribute information may include the following steps:
- Step 601 Determine the current Morton code of the point to be predicted in the point cloud to be decoded.
- the decoder may first determine the current Morton code of the point to be predicted in the point cloud to be decoded.
- the geometric information of the point cloud to be decoded and the attribute information corresponding to the point cloud to be decoded are decoded separately.
- the geometric information is decoded, the geometric information is reconstructed, and the decoding of the attribute information will depend on the reconstructed geometric information.
- the decoder can use the set information of each point in the point cloud to be decoded to determine the Morton code of each point. After obtaining the Morton code of each point, the decoder can sort all the points in the point cloud to be decoded in ascending order to generate the corresponding Morton code of all points to be decoded. Point cloud sequence collection. Among them, all Morton codes of all points in the point cloud to be decoded are stored in the point cloud sequence set.
- the decoder when the decoder decodes the attribute information of the point cloud to be decoded, it can predict the attribute information of all the points of the point cloud to be decoded in order of Morton code from small to large. .
- the point to be predicted can be any point in the point cloud to be decoded, and the current Morton code is the Morton code of the point to be predicted, and the attribute information of the point to be predicted is predicted At this time, other points corresponding to other Morton codes smaller than the current Morton code have completed the attribute information prediction, and the corresponding attribute reconstruction information is generated.
- the decoder when the decoder predicts the attribute information of the point to be predicted, it needs to determine the Morton code of the point to be predicted first to obtain the current Morton code.
- Step 602 Based on the current Morton code, determine the target Morton code corresponding to the point to be predicted according to the preset neighbor information table; wherein, the preset neighbor information table is used to compare the neighboring position of the point to be predicted and the preset reference point. The coordinate difference between is determined.
- the decoder may determine the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table.
- the target Morton code can represent the Morton code corresponding to the neighboring position around the point to be predicted.
- the preset neighbor information table can represent the coordinate difference between the neighboring position around the point to be predicted and the preset reference point.
- the adjacent position of the point to be predicted may include coplanar adjacent points that are coplanar adjacent to the point to be predicted, and may also include collinear adjacent points that are collinearly adjacent to the point to be predicted.
- the preset reference point when predicting the attribute information of the point to be predicted, if the point to be predicted is taken as the spatial center of the point cloud to be decoded, then the preset reference point may be the point to be predicted as The coordinate origin of the center of the space. Specifically, based on the same preset reference point, the difference between the distance between the preset reference point and the point to be predicted and the distance between the preset reference point and the adjacent position of the point to be predicted is calculated to obtain The coordinate difference between the adjacent position and the point to be predicted.
- the preset reference point may be the point with the smallest Morton code in the point cloud to be decoded.
- the decoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code and according to the preset neighbor information table, it needs to first establish the preset phase corresponding to the point to be predicted. Neighbor information table. In other words, the preset neighbor information table is established based on the points to be predicted.
- the decoder when the decoder establishes the preset neighbor information table corresponding to the point to be predicted, it can operate on the current Morton code (cur_pos) of the point to be predicted. Specifically, the decoder You can first locate the Morton code (base_pos) of the preset reference point, and then calculate the coordinate difference between the preset reference point and the adjacent position of the point to be predicted based on the Morton code of the preset reference point, and then continue Morton coding is performed on these coordinate differences, so that a preset neighbor information table corresponding to the points to be predicted can be established.
- base_pos the decoder You can first locate the Morton code (base_pos) of the preset reference point, and then calculate the coordinate difference between the preset reference point and the adjacent position of the point to be predicted based on the Morton code of the preset reference point, and then continue Morton coding is performed on these coordinate differences, so that a preset neighbor information table corresponding to the points to be predicted can be established.
- the adjacent positions of the points to be predicted may include coplanar adjacent points that are coplanar adjacent to the point to be predicted, and collinear adjacent points that are coplanar and adjacent to the point to be predicted. Therefore, when the decoder establishes the preset neighbor information table corresponding to the point to be predicted, it can respectively calculate the first coordinate difference between the preset reference point and the coplanar adjacent points, and calculate the preset reference point and the collinear phase. For the second coordinate difference between adjacent points, Morton encoding can be performed on the first coordinate difference to obtain the coplanar Morton code corresponding to the coplanar adjacent points.
- Morton encoding can be performed on the second coordinate difference to obtain The collinear Morton code corresponding to the collinear adjacent points.
- the decoder can establish a preset neighbor information table corresponding to the point to be predicted based on the coplanar Morton code and the collinear Morton code.
- the decoder may determine the to-be-predicted based on the preset neighbor information table corresponding to the current Morton code The target Morton code corresponding to the point, where the target Morton code can represent the Morton code corresponding to the neighboring position around the point to be predicted.
- the decoder when the decoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table, since the preset neighbor information table stores the target Morton code, Morton code after Morton decoding is performed on the coordinate difference between the preset reference point and the adjacent position. Therefore, the decoder can extract the Morton code of the coordinate difference corresponding to the adjacent position from the preset adjacent information table. Code, the Morton code of the coordinate difference can be further used to determine the target Morton code.
- the decoder is performing the target Morton code
- the decoder is performing target Morton code
- the adjacent type first, where the adjacent type may include coplanar adjacent and collinear adjacent, and then based on the determined adjacent type, the common adjacent information table can be further extracted. Planar Morton code or collinear Morton code, and determine the target Morton code based on coplanar Morton code or collinear Morton code.
- the distance between the coplanar adjacent point and it is 1, then the distance between the coplanar adjacent point and the point is Based on the principle that the closer the distance, the stronger the correlation, the adjacent types of coplanar adjacent can be selected first, and then the coplanar Morton code can be extracted from the preset adjacent information table, so as to use the coplanar Morton code to determine the target model. Pause.
- the target Morton code can represent the Morton code corresponding to the neighboring position around the point to be predicted. Therefore, the decoder can use the target Morton code to treat the neighboring neighboring points around the prediction point. Click to find and confirm.
- Step 603 Determine whether there are adjacent points of the point to be predicted in the point cloud to be decoded according to the target Morton code.
- the decoder after the decoder determines the target Morton code corresponding to the point to be predicted based on the current Morton code according to the preset neighbor information table, it can determine the point cloud to be decoded according to the target Morton code Whether there are adjacent points to the point to be predicted. In other words, the decoder can use Morton codes corresponding to adjacent positions around the point to be predicted to further determine whether there is a point adjacent to the point to be predicted in the decoded point cloud.
- the decoder after the decoder determines the target Morton code, it cannot directly determine the neighboring points of the point to be predicted according to the target Morton code. This is because the point cloud is parallel in space. It is not evenly distributed, and there may not really be adjacent points in the adjacent positions of the points to be predicted, that is to say, the target Morton code determined according to the adjacent positions of the points to be predicted may not really be in the point cloud to be decoded Therefore, after determining the target Morton code, the decoder also needs to determine whether there are adjacent points of the point to be predicted in the point cloud to be decoded according to the target Morton code.
- the decoder after obtaining the Morton code of each point, the decoder will sort all Morton codes of all points in ascending order to generate the points to be decoded.
- the point cloud sequence set corresponding to the cloud Therefore, after the decoder determines the target Morton code, it can use the point cloud sequence set corresponding to the point cloud to be decoded, combined with the target Morton code, to further determine whether there are adjacent points to be predicted Click to judge.
- the decoder may first obtain the point cloud of the point cloud to be decoded when judging whether there are adjacent points of the point to be predicted in the decoded point cloud according to the point cloud sequence set and the target Morton code. Sequence collection, and then use the target Morton code to perform query processing in the point cloud sequence collection, and the query result can be obtained. If the query result is that the target Morton code exists in the point cloud sequence collection, then the decoder can determine the point cloud to be decoded There are adjacent points corresponding to the points to be predicted. That is to say, in this application, after the decoder determines the target Morton code, it directly searches for the existence of the target Morton code in the point cloud sequence set. If it exists, it can be considered that the point to be predicted has adjacent points in space. ; Correspondingly, if it does not exist, it can be considered that the point to be predicted does not have an adjacent point in the space.
- the decoder when the decoder performs query processing in the point cloud sequence set based on the target Morton code, it can directly traverse each query point cloud sequence set according to the target Morton code. Morton code to obtain the query result; according to the target Morton code, the query processing of the point cloud sequence set can also be performed according to the preset query algorithm to obtain the query result.
- the decoder when the decoder performs query processing in the point cloud sequence set based on the target Morton code, it can either use the traversal method used for the point cloud sequence set or other point cloud sequence sets. Cloud sequence collection for fast search algorithm.
- Step 604 If there are adjacent points in the point cloud to be decoded, predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point.
- the decoder judges whether there are adjacent points of the point to be predicted in the decoded point cloud according to the target Morton code, if there are adjacent points of the point to be predicted in the point cloud to be decoded, the decoder The information can be directly reconstructed based on the attribute information of the adjacent points of the point to be predicted, and the attribute information of the point to be predicted can be predicted to obtain the attribute prediction value.
- the decoder may first obtain the adjacent point when predicting the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point. Then calculate the mean value of the reconstructed attribute value to obtain the mean value corresponding to the reconstructed attribute value, and then the mean value can be determined as the predictive value of the attribute.
- the adjacent points of the points to be predicted may be coplanar adjacent points that are coplanar and adjacent to the point to be predicted, or they may be collinearly adjacent to the point to be predicted. Collinear adjacent points. Therefore, if the decoder determines that the adjacent points of the points to be predicted are coplanar adjacent points or collinear adjacent points, then the decoder can obtain the co-linear adjacent points when obtaining the attribute reconstruction values of the adjacent points. The reconstructed value of the first attribute corresponding to the neighboring points of the plane, or obtaining the reconstructed value of the second attribute corresponding to the collinear neighbors.
- the decoder is performing the reconstruction of the attribute value.
- the first average value corresponding to at least one first reconstruction value can be calculated, or the second average value corresponding to at least one second reconstruction value can be calculated.
- the decoder can calculate the first average value or the second average value. The value is determined as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the decoder determines that the adjacent points of the points to be predicted are coplanar adjacent points and collinear adjacent points, then when the decoder obtains the attribute reconstruction values of the adjacent points, The first attribute reconstruction value corresponding to the coplanar adjacent points can be obtained, and at the same time, the second attribute reconstruction value corresponding to the collinear adjacent points can be obtained.
- the decoder since the point to be predicted has at least one coplanar adjacent coplanar adjacent point, and at least one collinear adjacent collinear adjacent point, the decoder is performing the reconstruction of the attribute value.
- a weighted average calculation can be performed on the reconstructed value of the first attribute and the reconstructed value of the second attribute according to the preset weight, so that the weighted average can be obtained.
- the decoder can determine the weighted average value as the attribute prediction value, thereby completing the prediction of the attribute information of the point to be predicted.
- the decoder when the decoder uses a preset weight to perform a weighted average calculation on the first attribute reconstruction value and the second attribute reconstruction value, it can separately calculate the first attribute reconstruction value and the second attribute reconstruction value.
- the reconstruction values are assigned different weight values. Specifically, based on the principle that the closer the distance, the stronger the correlation, the weight value corresponding to the first attribute reconstruction value should be greater than the weight value corresponding to the second attribute reconstruction value.
- the point to be predicted has multiple adjacent points such as coplanar adjacent points and collinear adjacent points, for example, there are 6 adjacent points to predict the attribute information of the point to be predicted, including 3 coplanar adjacent points and 3 collinear adjacent points.
- the decoder can perform a distance-based weighted average operation on the attribute reconstruction values of all adjacent points, so as to realize the prediction of the attribute information of the points to be predicted and obtain the attribute prediction values of the points to be predicted. That is to say, according to the distance between each adjacent point and the point to be predicted, the correlation decreases as the distance increases, so the attribute reconstruction value of the adjacent point with a smaller distance is given greater weight.
- the decoder when the decoder predicts the attribute information of the point to be predicted, it can simultaneously use the first attribute reconstruction value and the collinear phase of the coplanar adjacent points of the point to be predicted.
- the second attribute reconstruction value of the neighboring point determines the attribute prediction value of the point to be predicted.
- the attribute predicted value of the point to be predicted Among them, based on the principle that the closer the distance, the stronger the correlation, the closer coplanar neighboring points can be preferentially selected to predict the attribute information of the points to be predicted.
- the decoder may first search for coplanar adjacent points with a distance of 1, and if it exists, use the average value of the attribute reconstruction values of these coplanar adjacent points as the attribute prediction value of the point to be predicted ; If there is no coplanar adjacent point of the point to be predicted in the point cloud to be decoded, then continue to search for a distance of Collinear adjacent points, and then use the average value of the attribute reconstruction values of these collinear adjacent points as the attribute prediction value of the point to be predicted.
- FIG. 16 is a schematic diagram of the implementation process of the attribute information prediction method 7.
- the decoder determines whether there are adjacent points in the decoded point cloud according to the target Morton code, that is, step After 603, the method for the encoder to predict the attribute information of the point to be predicted may further include the following steps:
- Step 605 If there are no adjacent points in the decoded point cloud, use the previous Morton code corresponding to the current Morton code to predict the attribute prediction value of the point to be predicted.
- the decoder determines whether there are adjacent points of the point to be predicted in the decoded point cloud according to the target Morton code, if there are no adjacent points of the point to be predicted in the point cloud to be decoded, then decode Therefore, the decoder needs to use the previous Morton code corresponding to the current Morton code to predict the attribute information of the point to be predicted, so as to obtain the attribute information of the point to be predicted. The predicted value of the attribute of the point.
- the decoder when the decoder uses the previous Morton code corresponding to the current Morton code to predict the attribute prediction value of the point to be predicted, it may first determine the previous Morton code from the point cloud to be decoded For the corresponding reconstructed point, the attribute reconstruction value of the reconstructed point can then be obtained, and the attribute reconstruction value of the reconstructed point can be determined as the attribute prediction value of the point to be predicted.
- the reconstructed Morton code corresponding to the previous Morton code is directly
- the attribute reconstruction value of the point is used as the attribute prediction value of the point to be predicted to complete the prediction of the attribute information of the point to be predicted.
- the decoder can first search for the current Morton code from the point cloud sequence set of the point cloud to be decoded The previous Morton code. Among them, all Morton codes of all points in the point cloud to be decoded are stored in the point cloud sequence set in ascending order.
- Fig. 17 is a schematic diagram of the eighth implementation process of the method for predicting attribute information.
- step After 603, and before predicting the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the neighboring points, that is, before step 604, the method for the decoder to predict the attribute information of the point to be predicted may further include the following steps:
- Step 606 Determine the reconstructed point corresponding to the previous Morton code from the point cloud to be decoded.
- Step 607 Determine the reconstructed point as the neighboring point of the point to be predicted.
- the decoder after the decoder judges whether there are adjacent points of the point to be predicted in the decoded point cloud according to the target Morton code, if there are no adjacent points of the point to be predicted in the point cloud to be decoded, that is, In other words, there are no coplanar adjacent points and collinear adjacent points around the point to be predicted, so the decoder can first determine the reconstructed point corresponding to the previous Morton code from the point cloud to be decoded.
- the decoder can determine the neighboring point of the point to be predicted by the reconstructed point. That is to say, in this application, based on the target Morton code, if there is no adjacent point of the point to be predicted in the point cloud to be decoded, then the reconstructed Morton code corresponding to the previous Morton code is directly The points are regarded as adjacent points of the points to be predicted, and then the method of step 604 is continued to complete the prediction of the attribute information of the points to be predicted.
- step 603 that is, after the decoder determines whether there are adjacent points in the point cloud to be decoded according to the target Morton code, if there are points to be predicted in the point cloud to be decoded
- the decoder can directly predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point; if the adjacent point of the point to be predicted does not exist in the point cloud to be decoded, the decoder can start from the point to be predicted.
- the adjacent points of the points can continue to predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent points.
- the decoder can first compare the current point cloud sequence set to the current point cloud.
- the reconstructed point corresponding to the previous Morton code of the frame is regarded as the neighboring point of the point to be predicted, and then the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the neighboring point.
- the information of the point to be predicted can be used.
- the current Morton code determines the target Morton code corresponding to the adjacent position around the point to be predicted. If there is an adjacent point to the point to be predicted in the adjacent position in the point cloud, the codec can directly use the phase.
- the attribute reconstruction value prediction of the neighboring points obtains the attribute prediction value of the point to be predicted. It is precisely because the correlation between the point to be predicted and the neighboring point is stronger, so the prediction of the attribute information of the point to be predicted can be achieved through the neighboring points. Effectively improve the accuracy of forecasts.
- the decoder can also apply the above-mentioned method from step 601 to step 607 to the intra-layer prediction of the point cloud to be decoded.
- the method for predicting attribute information proposed in the present application can make full use of the spatial correlation of the point cloud, so that the prediction residual is smaller, and it is more suitable for subsequent quantization and entropy decoding. Under the premise of PSNR performance, higher gain is achieved.
- the embodiment of the application provides a method for predicting attribute information.
- the decoder determines the current Morton code of the point to be predicted in the point cloud to be decoded; based on the current Morton code, the corresponding point to be predicted is determined according to the preset neighbor information table.
- the target Morton code wherein the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code, it is judged whether there is a point cloud to be decoded.
- the adjacent points of the predicted point; if there are adjacent points in the point cloud to be decoded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- FIG. 18 is a schematic diagram of the composition structure of the encoder.
- the encoder 100 proposed in the embodiment of the present application may include a first determination. Part 101, the first judgment part 102, and the first prediction part 103.
- the first determining part 101 is configured to determine the current Morton code of the point to be predicted in the point cloud to be coded; and based on the current Morton code, determine the corresponding to the point to be predicted according to a preset neighbor information table.
- Target Morton code wherein the preset neighbor information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the first judging part 102 is configured to judge whether there are adjacent points of the point to be predicted in the point cloud to be coded according to the target Morton code;
- the first prediction part 103 is configured to predict the attribute prediction value of the point to be predicted according to the attribute reconstruction information of the adjacent point if the adjacent point exists in the point cloud to be coded.
- FIG. 19 is a schematic diagram of the composition structure of the encoder.
- the encoder 100 proposed in the embodiment of the present application may further include a first processor 104 and a first memory 105 storing executable instructions of the first processor 104 , A first communication interface 106, and a first bus 107 for connecting the first processor 104, the first memory 105, and the first communication interface 106.
- the above-mentioned first processor 104 is configured to determine the current Morton code of the point to be predicted in the point cloud to be coded; based on the current Morton code, according to preset neighbor information Table to determine the target Morton code corresponding to the point to be predicted; wherein the preset neighbor information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the target Morton code it is judged whether there are adjacent points of the point to be predicted in the point cloud to be coded; if the adjacent points exist in the point cloud to be coded, then according to the The attribute reconstruction information predicts the attribute predicted value of the point to be predicted.
- the functional modules in this embodiment may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
- the above-mentioned integrated unit can be realized in the form of hardware or software function module.
- the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer readable storage medium.
- the technical solution of this embodiment is essentially or correct
- the part that the prior art contributes or all or part of the technical solution can be embodied in the form of a software product.
- the computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer).
- the aforementioned storage media include: U disk, mobile hard disk, read only memory (Read Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program codes.
- the embodiment of the application provides an encoder that determines the current Morton code of the point to be predicted in the point cloud to be encoded; based on the current Morton code, the target corresponding to the point to be predicted is determined according to a preset neighbor information table Morton code; among them, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; judge whether there is a point to be predicted in the point cloud to be coded according to the target Morton code If there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted based on the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- FIG. 20 is a schematic diagram 1 of the composition structure of the decoder.
- the decoder 200 proposed in this embodiment of the present application may include a second determining part 201.
- the second determining part 201 is configured to determine the current Morton code of the point to be predicted in the point cloud to be decoded; and based on the current Morton code, determine the corresponding to the point to be predicted according to a preset neighbor information table.
- Target Morton code wherein the preset neighbor information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the second judging part 202 is configured to judge whether there are adjacent points of the point to be predicted in the point cloud to be decoded according to the target Morton code;
- the second prediction part 203 is configured to predict the attribute prediction value of the point to be predicted based on the attribute reconstruction information of the adjacent point if the adjacent point exists in the point cloud to be decoded.
- FIG. 21 is a second schematic diagram of the structure of the decoder.
- the decoder 200 proposed in this embodiment of the present application may further include a second processor 204 and a second memory 205 storing executable instructions of the second processor 204 , A second communication interface 206, and a second bus 207 for connecting the second processor 204, the first memory 205, and the second communication interface 206.
- the above-mentioned second processor 204 is configured to determine the current Morton code of the point to be predicted in the point cloud to be decoded; based on the current Morton code, according to preset neighbor information
- the table determines the target Morton code corresponding to the point to be predicted; wherein the preset neighboring information table is used to determine the coordinate difference between the neighboring position of the point to be predicted and the preset reference point;
- the target Morton code judges whether there are adjacent points of the point to be predicted in the point cloud to be decoded; if the adjacent points exist in the point cloud to be decoded, according to the attributes of the adjacent points
- the reconstruction information predicts the predicted value of the attribute of the point to be predicted.
- the functional modules in this embodiment may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
- the above-mentioned integrated unit can be realized in the form of hardware or software function module.
- the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer readable storage medium.
- the technical solution of this embodiment is essentially or correct
- the part that the prior art contributes or all or part of the technical solution can be embodied in the form of a software product.
- the computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer).
- a computer, a server, or a network device, etc.) or a processor executes all or part of the steps of the method in this embodiment.
- the aforementioned storage media include: U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk and other media that can store program codes.
- the embodiment of the application provides a decoder that determines the current Morton code of the point to be predicted in the point cloud to be decoded; based on the current Morton code, the target corresponding to the point to be predicted is determined according to a preset neighbor information table Morton code; among them, the preset neighbor information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; judge whether there is a point to be predicted in the point cloud to be decoded according to the target Morton code If there are adjacent points in the point cloud to be decoded, the attribute prediction value of the point to be predicted is predicted based on the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
- the embodiments of the present application provide a computer-readable storage medium and a computer-readable storage medium, on which a program is stored, and when the program is executed by a processor, the method as described in the above-mentioned embodiment is implemented.
- the program instructions corresponding to a method for predicting attribute information in this embodiment can be stored on storage media such as optical disks, hard disks, and USB flash drives.
- storage media such as optical disks, hard disks, and USB flash drives.
- the target Morton code corresponding to the point to be predicted is determined according to a preset neighbor information table; wherein, the preset neighbor information table is used to determine the neighboring position of the point to be predicted Determine the coordinate difference with the preset reference point;
- the method further includes the following steps:
- the target Morton code corresponding to the point to be predicted is determined according to a preset neighbor information table; wherein, the preset neighbor information table is used to determine the neighboring position of the point to be predicted Determine the coordinate difference with the preset reference point;
- this application can be provided as methods, systems, or computer program products. Therefore, this application may adopt the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Moreover, this application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program codes.
- These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing equipment to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction device.
- the device realizes the functions specified in one or more processes in the schematic diagram and/or one block or more in the block diagram.
- These computer program instructions can also be loaded on a computer or other programmable data processing equipment, so that a series of operation steps are executed on the computer or other programmable equipment to produce computer-implemented processing, so as to execute on the computer or other programmable equipment.
- the instructions provide steps for implementing functions specified in one or more processes in the schematic diagram and/or one block or more in the block diagram.
- the embodiment of the application provides a method for predicting attribute information, an encoder, a decoder, and a storage medium.
- the encoder determines the current Morton code of the point to be predicted in the point cloud to be encoded; based on the current Morton code, according to the preset
- the adjacent information table determines the target Morton code corresponding to the point to be predicted; the preset adjacent information table is used to determine the coordinate difference between the adjacent position of the point to be predicted and the preset reference point; according to the target Morton code It is judged whether there are adjacent points of the point to be predicted in the point cloud to be coded; if there are adjacent points in the point cloud to be coded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the decoder determines the current Morton code of the point to be predicted in the point cloud to be decoded; based on the current Morton code, determines the target Morton code corresponding to the point to be predicted according to the preset neighbor information table; judges the target Morton code to be decoded according to the target Morton code Whether there are adjacent points of the point to be predicted in the point cloud; if there are adjacent points in the point cloud to be decoded, the attribute prediction value of the point to be predicted is predicted according to the attribute reconstruction information of the adjacent points.
- the codec predicts the attribute information of the point to be predicted, based on the preset neighbor information table, the current Morton code of the point to be predicted can be used to determine the surrounding area of the point to be predicted.
- the target Morton code corresponding to the adjacent position of the point cloud.
- the codec can directly use the attribute reconstruction value of the adjacent point to predict and obtain the to-be-predicted
- the attribute prediction value of a point is precisely because the correlation between the point to be predicted and the adjacent point is stronger, so the prediction of the attribute information of the point to be predicted through the adjacent point can effectively improve the accuracy of the prediction and greatly reduce The prediction residual is reduced, thereby improving the coding efficiency.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
Abstract
Description
Claims (39)
- 一种属性信息的预测方法,应用于编码器中,所述方法包括:确定待编码点云中的待预测点的当前莫顿码;基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码;其中,所述预设相邻信息表用于对所述待预测点的相邻位置与预设基准点之间的坐标差进行确定;根据所述目标莫顿码判断所述待编码点云中是否存在所述待预测点的相邻点;若所述待编码点云中存在所述相邻点,则根据所述相邻点的属性重建信息预测所述待预测点的属性预测值。
- 根据权利要求1所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码之前,所述方法还包括:建立所述待预测点对应的所述预设相邻信息表。
- 根据权利要求2所述的方法,其中,所述相邻位置包括:与待预测点共面相邻的共面相邻点,和与待预测点共线相邻的共线相邻点,所述建立所述待预测点对应的所述预设相邻信息表,包括:定位所述预设基准点;计算所述预设基准点与所述共面相邻点之间的第一坐标差,计算所述预设基准点与所述共线相邻点之间的第二坐标差;对所述第一坐标差进行莫顿编码,获得所述共面相邻点对应的共面莫顿码,对所述第二坐标差进行莫顿编码,获得所述共线相邻点对应的共线莫顿码;基于所述共面莫顿码和所述共线莫顿码,建立所述待预测点对应的所述预设相邻信息表。
- 根据权利要求3所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:从所述预设相邻信息表中读取所述待预测点对应的所述共面莫顿码和所述共线莫顿码;对所述共面莫顿码和所述预设基准点的莫顿码进行求和运算,获得第一和值,对所述共线莫顿码和所述预设基准点的莫顿码进行求和运算,获得第二和值;将所述第一和值和所述第二和值,确定为所述目标莫顿码。
- 根据权利要求3所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码之前,所述方法还包括:确定相邻类型;其中,所述相邻类型包括:共面类型和共线类型。
- 根据权利要求5所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:若所述相邻类型为所述共面类型,则从所述预设相邻信息表中读取所述待预测点对应的所述共面莫顿码;对所述共面莫顿码和所述预设基准点的莫顿码进行求和运算,获得第一和值;将所述第一和值确定为所述目标莫顿码。
- 根据权利要求5所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:若所述相邻类型为所述共线类型,则从所述预设相邻信息表中读取所述待预测点对应的所述共线莫顿码;对所述共线莫顿码和所述预设基准点的莫顿码进行求和运算,获得第二和值;将所述第二和值确定为所述目标莫顿码。
- 根据权利要求1所述的方法,其中,所述根据所述目标莫顿码判断所述待编码点云中是否存在所述待预测点的相邻点,包括:获取所述待编码点云的点云序列集合;基于所述目标莫顿码在所述点云序列集合中进行查询处理,获得查询结果;若所述查询结果为所述点云序列集合中存在所述目标莫顿码,则判定存在所述相邻点。
- 根据权利要求8所述的方法,其中,所述基于所述目标莫顿码在所述点云序列集合中进行查询处理,获得查询结果,包括:基于所述目标莫顿码,遍历所述点云序列集合,获得所述查询结果。
- 根据权利要求8所述的方法,其中,所述基于所述目标莫顿码在所述点云序列中进行查询处理,获得查询结果,包括:基于所述目标莫顿码,按照预设查询算法在所述点云序列集合中查询,获得所述查询结果。
- 根据权利要求1所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共面相邻的共面相邻点,则获取所述共面相邻点对应的第一属性重建值;计算所述第一属性重建值的第一平均值,并将所述第一平均值确定为所述属性预测值。
- 根据权利要求1所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共线相邻的共线相邻点,则获取所述共线相邻点对应的第二属性重建值;计算所述第二属性重建值的第二平均值,并将所述第二平均值确定为所述属性预测值。
- 根据权利要求1所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共面相邻的共面相邻点,和与待预测点共线相邻的共线相邻点,则获取所述共面 相邻点对应的第一属性重建值,和所述共线相邻点对应的第二属性重建值;按照预设权重对所述第一属性重建值和所述第二属性重建值进行加权平均计算,获得加权平均值;将所述加权平均值确定为所述属性预测值。
- 根据权利要求1所述的方法,其中,所述根据所述目标莫顿码判断所述待编码点云中是否存在所述相邻点之后,所述方法还包括:若所述待编码点云中不存在所述相邻点,则利用所述当前莫顿码对应的前一个莫顿码预测所述待预测点的属性预测值。
- 根据权利要求14所述的方法,其中,所述利用所述当前莫顿码对应的前一个莫顿码预测所述待预测点的属性预测值,包括:从所述待编码点云中确定所述前一个莫顿码对应的已重建点;将所述已重建点的重建值确定为所述属性预测值。
- 根据权利要求1所述的方法,其中,所述根据所述目标莫顿码判断所述待编码点云中是否存在所述相邻点之后,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值之前,所述方法还包括:从所述待编码点云中确定所述前一个莫顿码对应的已重建点;将所述已重建点确定为所述待预测点的所述相邻点。
- 根据权利要求1至16任一项所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值之后,所述方法还包括:根据所述属性预测值计算所述待预测点的预测残差;对所述预测残差进行编码,生成码流。
- 一种属性信息的预测方法,应用于解码器中,所述方法包括:确定待解码点云中的待预测点的当前莫顿码;基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码;其中,所述预设相邻信息表用于对所述待预测点的相邻位置与预设基准点之间的坐标差进行确定;根据所述目标莫顿码判断所述待解码点云中是否存在所述待预测点的相邻点;若所述待解码点云中存在所述相邻点,则根据所述相邻点的属性重建信息预测所述待预测点的属性预测值。
- 根据权利要求18所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码之前,所述方法还包括:建立所述待预测点对应的所述预设相邻信息表。
- 根据权利要求19所述的方法,其中,所述相邻位置包括:与待预测点共面相邻的共面相邻点,和与待预测点共线相邻的共线相邻点,所述建立所述待预测点对应的所述预设相邻信息表,包括:定位所述预设基准点;计算所述预设基准点与所述共面相邻点之间的第一坐标差,计算所述预设基准点与所述共线相邻点之间的第二坐标差;对所述第一坐标差进行莫顿编码,获得所述共面相邻点对应的共面莫顿码,对所述第二坐标差进行莫顿编码,获得所述共线相邻点对应的共线莫顿码;基于所述共面莫顿码和所述共线莫顿码,建立所述待预测点对应的所述预设相邻信息表。
- 根据权利要求20所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:从所述预设相邻信息表中读取所述待预测点对应的所述共面莫顿码和所述共线莫顿码;对所述共面莫顿码和所述预设基准点的莫顿码进行求和运算,获得第一和值,对所述共线莫顿码和所述预设基准点的莫顿码进行求和运算,获得第二和值;将所述第一和值和所述第二和值,确定为所述目标莫顿码。
- 根据权利要求20所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码之前,所述方法还包括:确定相邻类型;其中,所述相邻类型包括:共面类型和共线类型。
- 根据权利要求22所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:若所述相邻类型为所述共面类型,则从所述预设相邻信息表中读取所述待预测点对应的所述共面莫顿码;对所述共面莫顿码和所述预设基准点的莫顿码进行求和运算,获得第一和值;将所述第一和值确定为所述目标莫顿码。
- 根据权利要求22所述的方法,其中,所述基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码,包括:若所述相邻类型为所述共线类型,则从所述预设相邻信息表中读取所述待预测点对应的所述共线莫顿码;对所述共线莫顿码和所述预设基准点的莫顿码进行求和运算,获得第二和值;将所述第二和值确定为所述目标莫顿码。
- 根据权利要求18所述的方法,其中,所述根据所述目标莫顿码判断所述待解码点云中是否存在所述待预测点的相邻点,包括:获取所述待解码点云的点云序列集合;基于所述目标莫顿码在所述点云序列集合中进行查询处理,获得查询结果;若所述查询结果为所述点云序列集合中存在所述目标莫顿码,则判定存在所述相邻点。
- 根据权利要求25所述的方法,其中,所述基于所述目标莫顿码在所述点云序列集合中进行查询处理,获得查询结果,包括:基于所述目标莫顿码,遍历所述点云序列集合,获得所述查询结果。
- 根据权利要求25所述的方法,其中,所述基于所述目标莫顿码在所述点云序列中进行查询处理,获得查询结果,包括:基于所述目标莫顿码,按照预设查询算法在所述点云序列集合中查询,获得所述查询结果。
- 根据权利要求18所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共面相邻的共面相邻点,则获取所述共面相邻点对应的第一属性重建值;计算所述第一属性重建值的第一平均值,并将所述第一平均值确定为所述属性预测值。
- 根据权利要求18所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共线相邻的共线相邻点,则获取所述共线相邻点对应的第二属性重建值;计算所述第二属性重建值的第二平均值,并将所述第二平均值确定为所述属性预测值。
- 根据权利要求18所述的方法,其中,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值,包括:若所述相邻点是与待预测点共面相邻的共面相邻点,和与待预测点共线相邻的共线相邻点,则获取所述共面相邻点对应的第一属性重建值,和所述共线相邻点对应的第二属性重建值;按照预设权重对所述第一属性重建值和所述第二属性重建值进行加权平均计算,获得加权平均值;将所述加权平均值确定为所述属性预测值。
- 根据权利要求18所述的方法,其中,所述根据所述目标莫顿码判断所述待解码点云中是否存在所述相邻点之后,所述方法还包括:若所述待解码点云中不存在所述相邻点,则利用所述当前莫顿码对应的前一个莫顿码预测所述待预测点的属性预测值。
- 根据权利要求31所述的方法,其中,所述利用所述当前莫顿码对应的前一个莫顿码预测所述待预测点的属性预测值,包括:从所述待解码点云中确定所述前一个莫顿码对应的已重建点;将所述已重建点的重建值确定为所述属性预测值。
- 根据权利要求18所述的方法,其中,所述根据所述目标莫顿码判断所述待解码点云中是否存在所述相邻点之后,所述根据所述相邻点的属性重建信息预测所述待预测点的属性预测值之前,所述方法还包括:从所述待解码点云中确定所述前一个莫顿码对应的已重建点;将所述已重建点确定为所述待预测点的所述相邻点。
- 一种编码器,所述编码器包括:第一确定部分,第一判断部分,第一预测部分,所述第一确定部分,配置为确定待编码点云中的待预测点的当前莫顿码;以及基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码;其中,所述预设相邻信息表用于对所述待预测点的相邻位置与预设基准点之间的坐标差进行确定;所述第一判断部分,配置为根据所述目标莫顿码判断所述待编码点云中是否存在所述待预测点的相邻点;所述第一预测部分,配置为若所述待编码点云中存在所述相邻点,则根据所述相邻点的属性重建信息预测所述待预测点的属性预测值。
- 一种解码器,所述解码器包括:第二确定部分,第二判断部分,第二预测部分,所述第二确定部分,配置为确定待解码点云中的待预测点的当前莫顿码;以及基于所述当前莫顿码,根据预设相邻信息表确定所述待预测点对应的目标莫顿码;其中,所述预设相邻信息表用于对所述待预测点的相邻位置与预设基准点之间的坐标差进行确定;所述第二判断部分,配置为根据所述目标莫顿码判断所述待解码点云中是否存在所述待预测点的相邻点;所述第二预测部分,配置为若所述待解码点云中存在所述相邻点,则根据所述相邻点的属性重建信息预测所述待预测点的属性预测值。
- 一种编码器,所述编码器包括第一处理器、存储有所述第一处理器可执行指令的第一存储器,当所述指令被所述第一处理器执行时,实现如权利要求1-17任一项所述的方法。
- 一种解码器,所述解码器包括第二处理器、存储有所述第二处理器可执行指令的第二存储器,当所述指令被所述第二处理器执行时,实现如权利要求18-33任一项所述的方法。
- 一种计算机可读存储介质,其上存储有程序,应用于编码器中,所述程序被处理器执行时,实现如权利要求1-17任一项所述的方法。
- 一种计算机可读存储介质,其上存储有程序,应用于解码器中,所述程序被处理器执行时,实现如权利要求18-33任一项所述的方法。
Priority Applications (9)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP19954841.3A EP4072135A4 (en) | 2019-12-02 | 2019-12-02 | METHOD FOR PREDICTING ATTRIBUTE INFORMATION, ENCODER, DECODER AND STORAGE MEDIA |
JP2022532617A JP7430792B2 (ja) | 2019-12-02 | 2019-12-02 | 属性情報の予測方法、エンコーダ、デコーダ及び記憶媒体 |
CN201980102312.4A CN114667732A (zh) | 2019-12-02 | 2019-12-02 | 属性信息的预测方法、编码器、解码器、及存储介质 |
CN202210909589.1A CN115190300A (zh) | 2019-12-02 | 2019-12-02 | 属性信息的预测方法、编码器、解码器、及存储介质 |
PCT/CN2019/122499 WO2021108969A1 (zh) | 2019-12-02 | 2019-12-02 | 属性信息的预测方法、编码器、解码器、及存储介质 |
KR1020227019876A KR20220110502A (ko) | 2019-12-02 | 2019-12-02 | 속성 정보의 예측 방법, 인코더, 디코더 및 저장매체 |
US17/805,093 US11704840B2 (en) | 2019-12-02 | 2022-06-02 | Attribute information prediction method, encoder, decoder and storage medium |
US18/300,484 US20230252684A1 (en) | 2019-12-02 | 2023-04-14 | Attribute information prediction method, encoder, decoder and storage medium |
JP2024010387A JP2024050705A (ja) | 2019-12-02 | 2024-01-26 | 属性情報の予測方法、エンコーダ、デコーダ及び記憶媒体 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2019/122499 WO2021108969A1 (zh) | 2019-12-02 | 2019-12-02 | 属性信息的预测方法、编码器、解码器、及存储介质 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/805,093 Continuation US11704840B2 (en) | 2019-12-02 | 2022-06-02 | Attribute information prediction method, encoder, decoder and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021108969A1 true WO2021108969A1 (zh) | 2021-06-10 |
Family
ID=76220940
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/122499 WO2021108969A1 (zh) | 2019-12-02 | 2019-12-02 | 属性信息的预测方法、编码器、解码器、及存储介质 |
Country Status (6)
Country | Link |
---|---|
US (2) | US11704840B2 (zh) |
EP (1) | EP4072135A4 (zh) |
JP (2) | JP7430792B2 (zh) |
KR (1) | KR20220110502A (zh) |
CN (2) | CN114667732A (zh) |
WO (1) | WO2021108969A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022257978A1 (zh) * | 2021-06-11 | 2022-12-15 | 维沃移动通信有限公司 | 点云编码方法、解码方法及装置 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115688004B (zh) * | 2022-11-08 | 2023-09-29 | 中国民用航空总局第二研究所 | 一种基于希尔伯特编码的目标属性确定方法、介质及设备 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108520470A (zh) * | 2017-02-28 | 2018-09-11 | 百度在线网络技术(北京)有限公司 | 用于生成用户属性信息的方法和装置 |
WO2019055772A1 (en) * | 2017-09-14 | 2019-03-21 | Apple Inc. | COMPRESSION OF CLOUD OF POINTS |
CN109815842A (zh) * | 2018-12-29 | 2019-05-28 | 上海依图网络科技有限公司 | 一种确定待识别对象的属性信息的方法及装置 |
CN110312132A (zh) * | 2019-03-11 | 2019-10-08 | 杭州海康威视数字技术股份有限公司 | 一种编解码方法、装置及其设备 |
CN110418135A (zh) * | 2019-08-05 | 2019-11-05 | 北京大学深圳研究生院 | 一种基于邻居的权重优化的点云帧内预测方法及设备 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB201303540D0 (en) * | 2013-02-27 | 2013-04-10 | Arithmetica Ltd | Image processing |
US10535161B2 (en) * | 2017-11-09 | 2020-01-14 | Samsung Electronics Co., Ltd. | Point cloud compression using non-orthogonal projection |
SG11202009891QA (en) | 2018-04-19 | 2020-11-27 | Panasonic Ip Corp America | Three-dimensional data encoding method, three-dimensional data decoding method, three-dimensional data encoding device, and three-dimensional data decoding device |
-
2019
- 2019-12-02 EP EP19954841.3A patent/EP4072135A4/en active Pending
- 2019-12-02 CN CN201980102312.4A patent/CN114667732A/zh active Pending
- 2019-12-02 KR KR1020227019876A patent/KR20220110502A/ko active Search and Examination
- 2019-12-02 JP JP2022532617A patent/JP7430792B2/ja active Active
- 2019-12-02 WO PCT/CN2019/122499 patent/WO2021108969A1/zh unknown
- 2019-12-02 CN CN202210909589.1A patent/CN115190300A/zh active Pending
-
2022
- 2022-06-02 US US17/805,093 patent/US11704840B2/en active Active
-
2023
- 2023-04-14 US US18/300,484 patent/US20230252684A1/en active Pending
-
2024
- 2024-01-26 JP JP2024010387A patent/JP2024050705A/ja active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108520470A (zh) * | 2017-02-28 | 2018-09-11 | 百度在线网络技术(北京)有限公司 | 用于生成用户属性信息的方法和装置 |
WO2019055772A1 (en) * | 2017-09-14 | 2019-03-21 | Apple Inc. | COMPRESSION OF CLOUD OF POINTS |
CN109815842A (zh) * | 2018-12-29 | 2019-05-28 | 上海依图网络科技有限公司 | 一种确定待识别对象的属性信息的方法及装置 |
CN110312132A (zh) * | 2019-03-11 | 2019-10-08 | 杭州海康威视数字技术股份有限公司 | 一种编解码方法、装置及其设备 |
CN110418135A (zh) * | 2019-08-05 | 2019-11-05 | 北京大学深圳研究生院 | 一种基于邻居的权重优化的点云帧内预测方法及设备 |
Non-Patent Citations (1)
Title |
---|
See also references of EP4072135A4 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022257978A1 (zh) * | 2021-06-11 | 2022-12-15 | 维沃移动通信有限公司 | 点云编码方法、解码方法及装置 |
Also Published As
Publication number | Publication date |
---|---|
US11704840B2 (en) | 2023-07-18 |
JP7430792B2 (ja) | 2024-02-13 |
EP4072135A4 (en) | 2023-06-21 |
US20220292723A1 (en) | 2022-09-15 |
KR20220110502A (ko) | 2022-08-08 |
US20230252684A1 (en) | 2023-08-10 |
JP2024050705A (ja) | 2024-04-10 |
CN115190300A (zh) | 2022-10-14 |
JP2023510081A (ja) | 2023-03-13 |
CN114667732A (zh) | 2022-06-24 |
EP4072135A1 (en) | 2022-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6732152B1 (ja) | 点群エンコーダ | |
JP7330306B2 (ja) | 変換方法、逆変換方法、エンコーダ、デコーダ及び記憶媒体 | |
CN112218079B (zh) | 一种基于空间顺序的点云分层方法、点云预测方法及设备 | |
KR20220127323A (ko) | 슬라이스별 트리수프 노드 크기 | |
CN110572655A (zh) | 一种基于邻居权重的参数选取和传递的点云属性编码和解码的方法及设备 | |
JP7386337B2 (ja) | 分割方法、符号器、復号器及びコンピュータ記憶媒体 | |
JP2024050705A (ja) | 属性情報の予測方法、エンコーダ、デコーダ及び記憶媒体 | |
WO2022042538A1 (zh) | 一种基于块的点云几何帧间预测方法和解码方法 | |
CN115606188A (zh) | 一种点云编解码方法、编码器、解码器及存储介质 | |
CN108989799B (zh) | 一种编码单元参考帧的选择方法、装置及电子设备 | |
WO2021062772A1 (zh) | 预测方法、编码器、解码器及计算机存储介质 | |
WO2022022013A1 (zh) | 模式选择方法、装置、计算机可读存储介质及电子设备 | |
JP2023523272A (ja) | 最近傍検索方法、エンコーダ、デコーダ、及び記憶媒体 | |
WO2021062732A1 (zh) | 变换方法、逆变换方法、编码器、解码器及存储介质 | |
WO2022120594A1 (zh) | 点云编解码方法、编码器、解码器及计算机存储介质 | |
CN114730474A (zh) | 点云处理方法、编码器、解码器及存储介质 | |
KR20220157490A (ko) | 인트라 예측 방법, 장치, 인코더, 디코더 및 저장 매체 | |
WO2022217611A1 (zh) | 编解码方法及相关设备、存储介质 | |
US20220343550A1 (en) | Partitioning method, encoder, decoder and computer storage medium | |
WO2021139796A1 (zh) | 莫顿码的构建方法、编码器、解码器、及存储介质 | |
CN117581549A (zh) | 帧内预测、编解码方法及装置、编解码器、设备、介质 | |
CN115412715A (zh) | 一种点云属性信息的预测编解码方法及装置 | |
JP2005301781A (ja) | 画像作成方法,画像符号化方法,画像復号方法,画像作成装置,画像符号化装置,画像復号装置,画像作成プログラム,画像符号化プログラム,画像復号プログラムおよびそれらのプログラム記録媒体 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19954841 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2022532617 Country of ref document: JP Kind code of ref document: A |
|
ENP | Entry into the national phase |
Ref document number: 20227019876 Country of ref document: KR Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2019954841 Country of ref document: EP Effective date: 20220704 |