WO2023025174A1

WO2023025174A1 - Methods and devices for point cloud prediction, point cloud encoding, and point cloud decoding

Info

Publication number: WO2023025174A1
Application number: PCT/CN2022/114386
Authority: WO
Inventors: 张伟; 王静; 李革; 高文
Original assignee: 西安电子科技大学; 鹏城实验室
Priority date: 2021-08-24
Filing date: 2022-08-24
Publication date: 2023-03-02
Also published as: CN115720272A

Abstract

Disclosed are methods and devices for point cloud prediction, point cloud encoding and point cloud decoding. The point cloud prediction method comprises: obtaining an encoded candidate point according to an encoded point in a point cloud; obtaining a geometric distance between a point to be encoded and the encoded candidate point according to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud; obtaining an attribute distance between the point to be encoded and the encoded candidate point according to encoded attribute information corresponding to the encoded candidate point and encoded attribute information corresponding to the point to be encoded; obtaining an encoded matching point according to the geometric distance and the attribute distance; and according to the encoded matching point, predicting an attribute to be encoded corresponding to the point to be encoded to obtain a prediction value. The present invention better takes into account a relationship between the attribute of the encoded point and the attribute of the point to be encoded, obtaining a more accurate prediction value, and improving a point cloud encoding effect.

Description

点云预测、点云编码、点云解码方法及设备Point cloud prediction, point cloud encoding, point cloud decoding method and device

技术领域technical field

本发明涉及点云处理计算领域，具体是涉及点云预测、点云编码、点云解码方法及设备。The invention relates to the field of point cloud processing and computing, in particular to methods and equipment for point cloud prediction, point cloud encoding, and point cloud decoding.

背景技术Background technique

现有技术采用如图2所示的方法对点云进行编码，具体过程为首先对点云的几何位置进行坐标转换，使点云全都包含在一个bounding box(包围盒)中。然后再进行量化，这一步量化主要起到缩放的作用，由于量化取整，使得一部分点的几何信息相同，根据参数来决定是否移除重复点，量化和移除重复点这一过程属于预处理过程。接下来，按照广度优先遍历的顺序对bounding box进行划分(八叉树/四叉树/二叉树)，对每个节点的占位码进行编码。在基于八叉树的几何码框架中，将包围盒依次划分得到子立方体，对非空的(包含点云中的点)的子立方体继续进行划分，直到划分得到的叶子结点为1x1x1的单位立方体时停止划分，其次对叶子节点中所包含的点数进行编码，最终完成几何八叉树的编码，生成几何码流。几何编码完成后，对几何信息进行重建，利用重建的已编码点所对应的几何信息来对待编码点所对应的属性信息进行编码。具体过程：对已编码属性信息进行空间转换，之后需要对重建点云进行属性插值，使得重建点云和原始点云的属性误差最小。在属性信息编码中分为两个分支：属性预测与属性变换。属性预测过程如下：首先对点云进行重排序，然后进行差分预测。其中重排序的方法有两种：莫顿重排序和Hilbert重排序。当前AVS编码框架中均采用Hilbert码对点云进行重排序。然后对排序之后的点云使用差分方式进行属性预测，若当前待编码点与前一个已编码点的几何信息相同，即为重复点，则利用重复点的重建属性值作为当前待编码点的属性预测值，否则对当前待编码点选择前Hilbert序的m个点作为邻居候选点，之后再依据邻居候选点对待编码点进行编码。The existing technology uses the method shown in Figure 2 to encode the point cloud. The specific process is to firstly perform coordinate transformation on the geometric position of the point cloud, so that all the point clouds are contained in a bounding box (bounding box). Then quantize. This step of quantization mainly plays the role of scaling. Due to the rounding of quantization, the geometric information of some points is the same. It is determined whether to remove duplicate points according to the parameters. The process of quantization and removal of duplicate points belongs to preprocessing. process. Next, divide the bounding box (octree/quadtree/binary tree) according to the order of breadth-first traversal, and encode the placeholder code of each node. In the octree-based geometric code framework, the bounding box is divided into sub-cubes in turn, and the sub-cubes that are not empty (including points in the point cloud) are continued to be divided until the leaf nodes obtained by the division are 1x1x1 units. Stop dividing the cube, and then encode the points contained in the leaf nodes, and finally complete the encoding of the geometric octree to generate a geometric code stream. After the geometric encoding is completed, the geometric information is reconstructed, and the attribute information corresponding to the point to be encoded is encoded by using the reconstructed geometric information corresponding to the encoded point. Specific process: Space conversion is performed on the encoded attribute information, and then attribute interpolation needs to be performed on the reconstructed point cloud to minimize the attribute error between the reconstructed point cloud and the original point cloud. There are two branches in attribute information coding: attribute prediction and attribute transformation. The attribute prediction process is as follows: first the point cloud is re-ranked and then differentially predicted. There are two reordering methods: Morton reordering and Hilbert reordering. In the current AVS coding framework, the Hilbert code is used to reorder the point cloud. Then use the differential method to predict the attribute of the sorted point cloud. If the geometric information of the current point to be encoded is the same as that of the previous encoded point, it is a repeated point, and the reconstructed attribute value of the repeated point is used as the attribute of the current point to be encoded. Otherwise, select the m points of the previous Hilbert order as neighbor candidate points for the current point to be encoded, and then encode the point to be encoded according to the neighbor candidate points.

综上所述，现有技术在对点云进行编码时，只依靠已编码点与待编码点之间的几何信息而忽视了两者之间已经编码的属性信息之间的联系，因此现有的点云编码方法降低了编码质量。To sum up, when the existing technology encodes the point cloud, it only relies on the geometric information between the encoded point and the point to be encoded and ignores the relationship between the encoded attribute information between the two. Therefore, the existing The point cloud encoding method degrades the encoding quality.

因此，现有技术还有待改进和提高。Therefore, the prior art still needs to be improved and improved.

发明内容Contents of the invention

为解决上述技术问题，本发明提供了点云预测、点云编码、点云解码方法及设备，解决了现有的点云编码方法降低了编码质量的问题。In order to solve the above technical problems, the present invention provides point cloud prediction, point cloud encoding, and point cloud decoding methods and equipment, which solves the problem that the existing point cloud encoding method reduces the encoding quality.

为实现上述目的，本发明采用了以下技术方案：To achieve the above object, the present invention adopts the following technical solutions:

第一方面，本发明提供一种点云预测方法，其中，包括：In a first aspect, the present invention provides a point cloud prediction method, including:

根据点云中的已编码点，得到所述已编码点中的已编码候选点；Obtaining encoded candidate points in the encoded points according to the encoded points in the point cloud;

依据点云中的待编码点在点云中的位置与所述已编码候选点在点云中的位置，得到所述待编码点与所述已编码候选点之间的几何距离；According to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud, the geometric distance between the point to be encoded and the encoded candidate point is obtained;

依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离；Obtaining an attribute distance between the point to be encoded and the candidate point to be encoded according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the to-be-encoded point;

依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点；Obtaining an encoded matching point among the encoded candidate points that matches the point to be encoded according to the geometric distance and the attribute distance;

依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。According to the encoded matching point, predict the attribute to be encoded corresponding to the point to be encoded to obtain a predicted value.

在一种实现方式中，所述根据点云中的已编码点，得到所述已编码点中的已编码候选点，包括方法一或方法二或方法三：In an implementation manner, the encoding candidate points in the encoded points are obtained according to the encoded points in the point cloud, including method one or method two or method three:

方法一：将所有所述已编码点作为所述已编码候选点；Method 1: using all the encoded points as the encoded candidate points;

方法二：将所有所述已编码点按照对应的希尔伯特码排序，根据所述希尔伯特码排序选择得到所述已编码候选点；Method 2: sorting all the encoded points according to the corresponding Hilbert codes, and selecting the encoded candidate points according to the sorting of the Hilbert codes;

方法三：将所有所述已编码点按照对应的莫顿码排序，根据所述莫顿码排序选择得到所述已编码候选点。Method 3: sort all the encoded points according to the corresponding Morton codes, and select the encoded candidate points according to the sorting of the Morton codes.

在一种实现方式中，所述依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离，包括：In an implementation manner, according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the to-be-encoded point, the to-be-encoded point and the encoded candidate point are obtained Attribute distances between, including:

依据所述已编码点中的所述已编码候选点所对应的已编码属性信息，得到所述已编码候选点所对应的已编码属性信息中的已编码第一属性；Obtaining the encoded first attribute in the encoded attribute information corresponding to the encoded candidate point according to the encoded attribute information corresponding to the encoded candidate point in the encoded point;

依据所述待编码点所对应的所述已编码属性信息，得到所述待编码点所对应的所述已编码属性信息中的所述待编码点所对应的已编码第一属性；According to the encoded attribute information corresponding to the to-be-encoded point, obtain the encoded first attribute corresponding to the to-be-encoded point in the encoded attribute information corresponding to the to-be-encoded point;

依据所述已编码候选点所对应的已编码第一属性和所述待编码点所对应的已编码第一属性，得到所述待编码点与所述已编码候选点之间的差值的函数值作为属性距离。According to the encoded first attribute corresponding to the encoded candidate point and the encoded first attribute corresponding to the to-be-encoded point, the function of obtaining the difference between the to-be-encoded point and the encoded candidate point Value as attribute distance.

在一种实现方式中，所述依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点，包括：In an implementation manner, the obtaining the encoded matching points among the encoded candidate points that match the points to be encoded according to the geometric distance and the attribute distance includes:

对所述待编码点与所述已编码点中的所述已编码候选点之间的所述几何距离以及所述待编码点与所述已编码候选点之间的属性距离进行加权计算，得到所述待编码点和所述已编码候选点之间的综合距离；performing weighted calculation on the geometric distance between the point to be coded and the coded candidate point in the coded point and the attribute distance between the point to be coded and the coded candidate point to obtain The integrated distance between the points to be encoded and the encoded candidate points;

依据所述综合距离，按照距离大小得到所述已编码匹配点。According to the comprehensive distance, the coded matching points are obtained according to the distance.

在一种实现方式中，所述依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值，包括：In an implementation manner, the predicting the attribute to be encoded corresponding to the point to be encoded according to the encoded matching point to obtain a predicted value includes:

依据所述已编码匹配点所对应的已编码属性信息，得到所述已编码匹配点所对应的已编码属性信息中的已编码第二属性；Obtaining the encoded second attribute in the encoded attribute information corresponding to the encoded matching point according to the encoded attribute information corresponding to the encoded matching point;

依据所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Predict the second attribute to be encoded corresponding to the point to be encoded according to the encoded second attribute corresponding to the encoded matching point to obtain a prediction value.

在一种实现方式中，所述依据所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行进行预测，得到预测值，包括方法一或方法二：In an implementation manner, according to the encoded second attribute corresponding to the encoded matching point, the second attribute to be encoded corresponding to the point to be encoded is predicted to obtain a predicted value, including Method 1 or Method 2:

方法一：依据所述待编码点与所述已编码匹配点之间的几何距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值；Method 1: According to the geometric distance between the point to be encoded and the encoded matching point and the encoded second attribute corresponding to the encoded matching point, the to-be-encoded point corresponding to the to-be-encoded point is The second attribute is predicted and the predicted value is obtained;

方法二：依据所述待编码点与所述已编码匹配点之间的综合距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Method 2: According to the comprehensive distance between the point to be coded and the coded matching point and the coded second attribute corresponding to the coded matching point, the to-be-coded point corresponding to the coded point The second attribute is predicted and the predicted value is obtained.

在一种实现方式中，所述依据所述待编码点与所述已编码匹配点之间的综合距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值，包括：In an implementation manner, the to-be-coded point is coded according to the comprehensive distance between the to-be-coded point and the coded matching point and the coded second attribute corresponding to the coded matching point The second attribute to be encoded corresponding to the point is predicted to obtain the predicted value, including:

依据所述综合距离，得到所述已编码第二属性所对应的权重；Obtaining the weight corresponding to the encoded second attribute according to the comprehensive distance;

依据所述已编码第二属性所对应的权重和所述已编码第二属性所对应的属性值，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Predict the second attribute to be encoded corresponding to the point to be encoded according to the weight corresponding to the encoded second attribute and the attribute value corresponding to the encoded second attribute, to obtain a predicted value.

第二方面，本发明实施例还提供一种点云编码方法，包括：In the second aspect, the embodiment of the present invention also provides a point cloud encoding method, including:

获取待编码点；Obtain the points to be encoded;

依据上述所述的点云预测方法，得到所述待编码点所对应的预测值；According to the point cloud prediction method described above, the predicted value corresponding to the point to be encoded is obtained;

获取所述待编码点所对应的待编码第二属性的设定值；Acquiring the set value of the second attribute to be encoded corresponding to the point to be encoded;

依据所述待编码第二属性的设定值和所述预测值，得到所述设定值与所述预测值之间的预测残差值；Obtaining a prediction residual value between the set value and the predicted value according to the set value of the second attribute to be encoded and the predicted value;

依据所述预测残差值，对所述待编码点所对应的待编码第二属性进行编码，得到码流。According to the prediction residual value, encode the second attribute to be encoded corresponding to the point to be encoded to obtain a code stream.

第三方面，本发明实施例还提供一种点云解码方法，包括：In a third aspect, the embodiment of the present invention also provides a point cloud decoding method, including:

获取待解码点所对应的码流；Obtain the code stream corresponding to the point to be decoded;

依据所述码流和上述所述的点云预测方法，得到所述待解码点所对应的预测值；Obtain the predicted value corresponding to the point to be decoded according to the code stream and the above-mentioned point cloud prediction method;

依据所述码流，得到待解码点对应的预测残差值According to the code stream, the prediction residual value corresponding to the point to be decoded is obtained

依据所述预测残差值和所述预测值，得到所述待解码点的第二属性重建值。Obtain a second attribute reconstruction value of the point to be decoded according to the prediction residual value and the prediction value.

第四方面，本发明实施例还提供一种点云预测装置，所述装置包括如下组成部分：In the fourth aspect, the embodiment of the present invention also provides a point cloud prediction device, which includes the following components:

已编码候选点确定模块，用于根据点云中的已编码点，得到所述已编码点中的已编码候选点；The coded candidate point determination module is used to obtain the coded candidate points in the coded points according to the coded points in the point cloud;

几何距离计算模块，用于依据点云中的待编码点在点云中的位置与所述已编码候选点在点云中的位置，得到所述待编码点与所述已编码候选点之间的几何距离；The geometric distance calculation module is used to obtain the distance between the point to be encoded and the encoded candidate point according to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud. the geometric distance;

属性距离计算模块，用于依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离；An attribute distance calculation module, configured to obtain the distance between the point to be encoded and the candidate point to be encoded according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the point to be encoded. attribute distance;

匹配点计算模型，用于依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点；A matching point calculation model, used to obtain an encoded matching point that matches the point to be encoded among the encoded candidate points according to the geometric distance and the attribute distance;

预测模块，用于依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。The prediction module is configured to predict the attribute to be encoded corresponding to the point to be encoded according to the encoded matching point, and obtain a predicted value.

第五方面，本发明实施例还提供一种终端设备，其中，所述终端设备包括存储器、处理器及存储在所述存储器中并可在所述处理器上运行的点云预测程序，所述处理器执行所述点云预测程序时，实现上述所述的点云预测方法的步骤。In the fifth aspect, an embodiment of the present invention further provides a terminal device, wherein the terminal device includes a memory, a processor, and a point cloud prediction program stored in the memory and operable on the processor, the When the processor executes the point cloud prediction program, the steps of the above-mentioned point cloud prediction method are realized.

第六方面，本发明实施例还提供一种计算机可读存储介质，所述计算机可读存储介质上存储有点云预测程序，所述点云预测程序被处理器执行时，实现上述所述的点云预测方法的步骤。In the sixth aspect, the embodiment of the present invention also provides a computer-readable storage medium, where a point cloud prediction program is stored on the computer-readable storage medium, and when the point cloud prediction program is executed by a processor, the above-mentioned points can be realized. Steps of cloud forecasting method.

有益效果：本发明在几何距离的基础上又增加了属性距离(属性之间的差异)，之后将几何距离和属性距离进行结合，得到点云中与待编码点所匹配的已编码点，只有找到合适的已编码点才能对待编码点进行准确的编码。本发明的属性距离是待编码点所对应的已编码属性与已编码点所对应的已编码属性之间的距离，这两个已编码属性之间存在相关性，而待编码点所对应的已编码属性和待编码点所对应的待编码属性又不同，从而实现了跨属性编码。本发明采用跨属性编码充分考虑了已编码点属性与待编码属性之间的联系，从而在基于几何距离的基础上结合了属性距离以降低待编码点与已编码点之间的属性残差值，进而提高点云编码质量和效率。Beneficial effects: the present invention increases the attribute distance (difference between attributes) on the basis of the geometric distance, then combines the geometric distance and the attribute distance to obtain the encoded point matched with the point to be encoded in the point cloud, only Only by finding the appropriate coded point can the coded point to be coded be accurately coded. The attribute distance in the present invention is the distance between the encoded attribute corresponding to the point to be encoded and the encoded attribute corresponding to the encoded point. The attributes to be encoded corresponding to the encoding attributes and the points to be encoded are different, so that cross-attribute encoding is realized. The present invention uses cross-attribute coding to fully consider the connection between the encoded point attribute and the attribute to be encoded, so that the attribute distance is combined on the basis of the geometric distance to reduce the residual value of the attribute between the point to be encoded and the encoded point , thus improving the quality and efficiency of point cloud encoding.

附图说明Description of drawings

图1为本发明的整体流程图；Fig. 1 is the overall flowchart of the present invention;

图2为背景技术中现有技术的流程图；Fig. 2 is the flow chart of prior art in background technology;

图3为本发明的具体实施例的流程图；Fig. 3 is the flowchart of the specific embodiment of the present invention;

图4为本发明中的实施例对颜色属性的预测结果；Fig. 4 is the prediction result of the embodiment in the present invention to color property;

图5为本发明中的实施例对折射率属性的预测结果。Fig. 5 is the prediction result of the embodiment of the present invention on the refractive index property.

具体实施方式Detailed ways

以下结合实施例和说明书附图，对本发明中的技术方案进行清楚、完整地描述。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the present invention are clearly and completely described below in conjunction with the embodiments and the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

经研究发现，现有技术采用如图2所示的方法对点云进行编码，具体过程为首先对点云的几何位置进行坐标转换，使点云全都包含在一个bounding box(包围盒)中。然后再进行量化，这一步量化主要起到缩放的作用，由于量化取整，使得一部分点的几何信息相同，根据参数来决定是否移除重复点，量化和移除重复点这一过程属于预处理过程。接下来，按照广度优先遍历的顺序对bounding box进行划分(八叉树/四叉树/二叉树)，对每个节点的占位码进行编码。在基于八叉树的几何码框架中，将包围盒依次划分得到子立方体，对非空的(包含点云中的点)的子立方体继续进行划分，直到划分得到的叶子结点为1x1x1的单位立方体时停止划分，其次对叶子节点中所包含的点数进行编码，最终完成几何八叉树的编码，生成几何码流。几何编码完成后，对几何信息进行重建，利用重建的已编码点所对应的几何信息来对待编码点所对应的属性信息进行编码。具体过程：对已编码属性信息进行空间转换，之后需要对重建点云进行属性插值，使得重建点云和原始点云的属性误差最小。在属性信息编码中分为两个分支：属性预测与属性变换。属性预测过程如下：首先对点云进行重排序，然后进行差分预测。其中重排序的方法有两种：莫顿重排序和Hilbert重排序。当前AVS编码框架中均采用Hilbert码对点云进行重排序。然后对排序之后的点云使用差分方式进行属性预测，若当前待编码点与前一个已编码点的几何信息相同，即为重复点，则利用重复点的重建属性值作为当前待编码点的属性预测值，否则对当前待编码点选择前Hilbert序的m个点作为邻居候选点，之后再依据邻居候选点对待编码点进行编码。现有技术在对点云进行编码时，只依靠已编码点与待编码点之间的几何信息而忽视了两者之间已经编码的属性信息之间的联系，因此现有的点云编码方法降低了编码质量。After research, it is found that the existing technology adopts the method shown in Figure 2 to encode the point cloud. The specific process is to firstly perform coordinate conversion on the geometric position of the point cloud, so that the point cloud is all contained in a bounding box (bounding box). Then quantize. This step of quantization mainly plays the role of scaling. Due to the rounding of quantization, the geometric information of some points is the same. It is determined whether to remove duplicate points according to the parameters. The process of quantization and removal of duplicate points belongs to preprocessing. process. Next, divide the bounding box (octree/quadtree/binary tree) according to the order of breadth-first traversal, and encode the placeholder code of each node. In the octree-based geometric code framework, the bounding box is divided into sub-cubes in turn, and the sub-cubes that are not empty (including points in the point cloud) are continued to be divided until the leaf nodes obtained by the division are 1x1x1 units. Stop dividing the cube, and then encode the points contained in the leaf nodes, and finally complete the encoding of the geometric octree to generate a geometric code stream. After the geometric encoding is completed, the geometric information is reconstructed, and the attribute information corresponding to the point to be encoded is encoded by using the reconstructed geometric information corresponding to the encoded point. Specific process: Space conversion is performed on the encoded attribute information, and then attribute interpolation needs to be performed on the reconstructed point cloud to minimize the attribute error between the reconstructed point cloud and the original point cloud. There are two branches in attribute information coding: attribute prediction and attribute transformation. The attribute prediction process is as follows: first the point cloud is re-ranked and then differentially predicted. There are two reordering methods: Morton reordering and Hilbert reordering. In the current AVS coding framework, the Hilbert code is used to reorder the point cloud. Then use the differential method to predict the attribute of the sorted point cloud. If the geometric information of the current point to be encoded is the same as that of the previous encoded point, it is a repeated point, and the reconstructed attribute value of the repeated point is used as the attribute of the current point to be encoded. Otherwise, select the m points of the previous Hilbert order as neighbor candidate points for the current point to be encoded, and then encode the point to be encoded according to the neighbor candidate points. When the existing technology encodes the point cloud, it only relies on the geometric information between the encoded point and the point to be encoded and ignores the relationship between the encoded attribute information between the two. Therefore, the existing point cloud encoding method Reduced encoding quality.

为解决上述技术问题，本发明提供了点云预测、点云编码、点云解码方法及设备，解决了现有的点云编码方法降低了编码质量的问题。具体实施时，本发明在几何距离的基础上又增加了属性距离，之后将几何距离和属性距离进行结合，以对待编码点中待编码属性进行编码。本发明采用跨属性编码充分考虑了已编码点属性与待编码属性之间的联系，从而在基于几何距离的基础上结合了属性距离以降低待编码点与已编码点之间的属性残差值，进而提高点云编码质量和效率。In order to solve the above technical problems, the present invention provides point cloud prediction, point cloud encoding, and point cloud decoding methods and equipment, which solves the problem that the existing point cloud encoding method reduces the encoding quality. During specific implementation, the present invention adds the attribute distance on the basis of the geometric distance, and then combines the geometric distance and the attribute distance to encode the attributes to be encoded in the points to be encoded. The present invention uses cross-attribute coding to fully consider the connection between the encoded point attribute and the attribute to be encoded, so that the attribute distance is combined on the basis of the geometric distance to reduce the residual value of the attribute between the point to be encoded and the encoded point , thus improving the quality and efficiency of point cloud encoding.

举例说明，点云中的所有点都至少具有两个属性A和B，已编码点中的属性A和属性B都已经编码了，待编码点中的属性A已经编码了，但是待编码中的属性B未编码。本实施例，首先计算待编码点与已编码点之间的几何距离(几何距离即两者在点云中的位置之间的距离，可以是欧氏距离或曼哈顿距离)，同时计算待编码点的属性A与已编码点中的属性A之间的属性距离，属性距离包括但不限于两者之间的差值，得到已编码点和待编码点之间的属性距离和几何距离之后，上述得到点云中的各个已编码点与待编码点之间的属性距离和几何距离之后，再依据属性距离和几何距离得到已编码点中与待编码点所匹配的已编码匹配点。比如有100个已编码点，但是通过计算这100个已编码点与待编码点之间的几何距离和属性距离之后，100个已编码点中只有5个已编码点符合要求。通过几何距离和属性距离得到的已编码匹配点才是与待编码点相关的点，才能提高待编码点的待编码属性的编码效率。For example, all points in the point cloud have at least two attributes A and B, attribute A and attribute B in the coded points have been coded, and attribute A in the point to be coded has been coded, but the coded point Property B is not encoded. In this embodiment, first calculate the geometric distance between the point to be encoded and the encoded point (the geometric distance is the distance between the positions of the two in the point cloud, which can be Euclidean distance or Manhattan distance), and calculate the point to be encoded at the same time The attribute distance between the attribute A of the coded point and the attribute A in the coded point, the attribute distance includes but not limited to the difference between the two, after obtaining the attribute distance and the geometric distance between the coded point and the point to be coded, the above After obtaining the attribute distance and geometric distance between each encoded point in the point cloud and the point to be encoded, the encoded matching point matching the point to be encoded among the encoded points is obtained according to the attribute distance and geometric distance. For example, there are 100 coded points, but after calculating the geometric distance and attribute distance between the 100 coded points and the points to be coded, only 5 of the 100 coded points meet the requirements. The encoded matching points obtained by geometric distance and attribute distance are the points related to the points to be encoded, so as to improve the encoding efficiency of the attributes to be encoded of the points to be encoded.

示例性方法exemplary method

本实施例的一种点云编码方法可应用于终端设备中。在本实施例中，如图1中所示，所述点云编码方法具体包括如下步骤：A point cloud encoding method in this embodiment can be applied to a terminal device. In the present embodiment, as shown in Figure 1, the point cloud encoding method specifically includes the following steps:

S100，依据点云中的待编码点在点云中的位置与已编码点在点云中的位置，得到所述待编码点与所述已编码点之间的几何距离。S100. Obtain the geometric distance between the point to be encoded and the encoded point according to the position of the point to be encoded in the point cloud and the position of the encoded point in the point cloud.

本实施例中已编码点为已编码候选点。In this embodiment, the coded points are coded candidate points.

点云中分布有很多点，这些点在点云中有相应的位置，不同位置的点之间的距离是不同的，而且点与点之间的距离与点所对应的属性有一定的联系。There are many points distributed in the point cloud. These points have corresponding positions in the point cloud. The distances between points in different positions are different, and the distance between points has a certain relationship with the attributes corresponding to the points.

步骤S100包括如下的步骤S101和S102：Step S100 includes the following steps S101 and S102:

S101，依据所述点云中的所有所述已编码点所对应的希尔伯特码和所述待编码点所对应的希尔伯特码，得到所述已编码点中与所述待编码点所匹配的已编码候选点。S101. According to the Hilbert codes corresponding to all the encoded points in the point cloud and the Hilbert codes corresponding to the to-be-encoded points, obtain the coded points corresponding to the to-be-encoded points The encoded candidate points that the point matches.

本实施例除了可以依据希尔伯特码得到已编码候选点，还可以根据莫顿码得到已编码候选点。或者将所有的已编码点都作为已编码候选点In this embodiment, besides the encoded candidate points can be obtained according to the Hilbert code, the encoded candidate points can also be obtained according to the Morton code. Or use all coded points as coded candidate points

首先对点云中的所有已编码点和待编码点按照希尔伯特码进行升序排列，得到排序之后的点云。再依据排序之后的所述点云，得到所述已编码点中的已编码候选点，所述已编码候选点为排序之后的所述点云中的位于所述待编码点之前设定个数的所述已编码点。First, all the encoded points and points to be encoded in the point cloud are sorted in ascending order according to the Hilbert code, and the sorted point cloud is obtained. According to the point cloud after sorting, the coded candidate points in the coded points are obtained, and the coded candidate points are the set number before the point to be coded in the point cloud after sorting The encoded points of .

S102，依据所述待编码点在所述点云中的位置和所述已编码候选点在所述点云中的位置，得到所述待编码点与所述已编码候选点之间的几何距离。S102. According to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud, obtain the geometric distance between the point to be encoded and the encoded candidate point .

所谓的几何距离就是在物理空间上点与点之间的距离，通过几何距离对点云中所有已编码点进行筛选，以找到待编码点最近邻，即在后续的对待编码的待编码属性的预测编码过程中使用点云的几何信息作为待编码点最近邻的判定条件，可以先筛选出一批与待编码点相关的已编码候选点。本实施例根据已编码点能够得到待编码点的待编码属性的预测值，再根据预测值对待编码属性进行编码。计算几何距离的具体过程如下：The so-called geometric distance is the distance between points in the physical space. All encoded points in the point cloud are screened through the geometric distance to find the nearest neighbor of the point to be encoded, that is, in the subsequent attribute to be encoded to be encoded In the process of predictive encoding, the geometric information of the point cloud is used as the judgment condition of the nearest neighbor of the point to be encoded, and a batch of encoded candidate points related to the point to be encoded can be screened out first. In this embodiment, the predicted value of the attribute to be encoded of the point to be encoded can be obtained according to the encoded point, and then the attribute to be encoded is encoded according to the predicted value. The specific process of calculating the geometric distance is as follows:

首先，对应当前已编码点，按照空间填充曲线(如Hilbert序，即希尔伯特码排序)排序，将前序m个点作为邻居候选点(已编码候选点)；然后遍历所有邻居候选点，计算各个邻居候选点与当前待编码点的几何距离(disGeom)。其中，计算公式如下所示：First, corresponding to the current coded point, sort according to the space filling curve (such as Hilbert order, that is, Hilbert code sorting), and use the previous m points as neighbor candidate points (coded candidate points); then traverse all neighbor candidate points , calculate the geometric distance (disGeom) between each neighbor candidate point and the current point to be encoded. Among them, the calculation formula is as follows:

disGeom＝dist(curGeom,neighbourGeom)disGeom = dist(curGeom, neighbourGeom)

disGeom为计算得到的已编码候选点到待编码点几何距离，curGeom表示当前待编码点的几何坐标，neighbourGeom表示邻居候选点的几何坐标。函数dist(curGeom，neighbourGeom)表示curGeom与neighbourGeom之间的距离，该距离可以为欧氏距离或曼哈顿距离。几何信息所表示的空间维度包含但不限于点云的笛卡尔坐标、球面坐标、柱面坐标等disGeom is the calculated geometric distance from the encoded candidate point to the point to be encoded, curGeom indicates the geometric coordinates of the current point to be encoded, and neighbourGeom indicates the geometric coordinates of the neighboring candidate points. The function dist(curGeom, neighbourGeom) indicates the distance between curGeom and neighbourGeom, which can be Euclidean distance or Manhattan distance. The spatial dimensions represented by geometric information include but are not limited to Cartesian coordinates, spherical coordinates, cylindrical coordinates, etc. of the point cloud

S200，依据所述已编码点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码点之间的属性距离，所述已编码点所对应的已编码属性信息与所述待编码点所对应的已编码属性信息所匹配。S200. According to the encoded attribute information corresponding to the encoded point and the encoded attribute information corresponding to the to-be-encoded point, obtain the attribute distance between the to-be-encoded point and the encoded point, and the The encoded attribute information corresponding to the encoding point matches the encoded attribute information corresponding to the point to be encoded.

本实施例中的属性距离可以是属性所对应的属性差值，但不仅仅只包括属性差值。步骤S200包括如下的步骤S201和步骤S202：The attribute distance in this embodiment may be the attribute difference corresponding to the attribute, but not only includes the attribute difference. Step S200 includes the following steps S201 and S202:

S201，依据所述已编码点中的所述已编码候选点所对应的已编码属性信息，得到所述已编码候选点所对应的已编码属性信息中的已编码第一属性。S201. Obtain an encoded first attribute in the encoded attribute information corresponding to the encoded candidate point according to the encoded attribute information corresponding to the encoded candidate point in the encoded point.

本实施例中，当已编码候选点只包含两个属性时，已编码候选点所对应的所有属性都是已经编码过的，当已编码候选点包括三个及三个以上属性时，其中第一个属性和第二个属性是用于对待编码点中的属性进行编码的，那么已编码候选点所对应的第一个属性和第二个属性是全部被编码过的，而不参与对待编码点中的属性进行编码的第三个属性可以是编码过的也可以是没有被编码过的，而待编码点中有的属性已经编码了，有的属性没有编码，本实施例用待编码点中的已编码属性去计算它与已编码点之间的属性距离。In this embodiment, when the encoded candidate points only contain two attributes, all the attributes corresponding to the encoded candidate points have been encoded; when the encoded candidate points include three or more attributes, the first One attribute and the second attribute are used to encode the attribute in the coded point, then the first attribute and the second attribute corresponding to the coded candidate point are all coded, and do not participate in the coded point The third attribute encoded by the attribute in the point can be encoded or not encoded, and some attributes in the points to be encoded have been encoded, and some attributes have not been encoded. In this embodiment, the points to be encoded are used The encoded attribute in to calculate the attribute distance between it and the encoded point.

S202，依据所述已编码候选点所对应的已编码第一属性和所述待编码点所对应的已编码第一属性，得到所述待编码点与所述已编码候选点之间的属性距离。S202. According to the encoded first attribute corresponding to the encoded candidate point and the encoded first attribute corresponding to the to-be-encoded point, obtain the attribute distance between the to-be-encoded point and the encoded candidate point .

如上述所述，可以用已编码第一属性与待编码点中的已编码第一属性之间的差异(差值)作为待编码点与已编码候选点之间的属性距离。As mentioned above, the difference (difference value) between the encoded first attribute and the encoded first attribute in the point to be encoded can be used as the attribute distance between the point to be encoded and the encoded candidate point.

本实施例中，计算属性距离时，待编码点所对应的已编码属性为已编码第一属性。In this embodiment, when calculating the attribute distance, the encoded attribute corresponding to the point to be encoded is the encoded first attribute.

计算属性距离的具体过程如下：The specific process of calculating attribute distance is as follows:

在遍历所有已编码候选点时，同时计算各个已编码候选点与当前待编码点的已编码属性信息之间的距离(disAttr)。其中属性距离的计算公式如下所示：When traversing all the encoded candidate points, the distance (disAttr) between each encoded candidate point and the encoded attribute information of the current point to be encoded is calculated at the same time. The formula for calculating attribute distance is as follows:

disAttr＝distance(curAttr,neighbourAttr)disAttr = distance(curAttr, neighbourAttr)

disAttr为计算得到的属性距离，curAttr表示当前待编码点的已编码属性信息，neighbourAttr表示已编码候选点的已编码第一属性。其中属性信息包含但不限于点云的颜色、反射率等。函数distance(curAttr，neighbourAttr)表示curAttr点与neighbourAttr点之间的属性距离(差异)函数，在本方案中，该距离(差异)的计算可以表示为如下所示：disAttr is the calculated attribute distance, curAttr indicates the encoded attribute information of the current point to be encoded, and neighborAttr indicates the encoded first attribute of the encoded candidate point. The attribute information includes but is not limited to the color, reflectivity, etc. of the point cloud. The function distance(curAttr, neighbourAttr) represents the attribute distance (difference) function between the curAttr point and the neighbourAttr point. In this solution, the calculation of the distance (difference) can be expressed as follows:

其中，attrDim表示属性的个数，例如反射率的分量个数为1，颜色的分量个数为3。curAttr表示当前待编码点的已编码属性信息，neighbourAttr表示已编码候选点的已编码第一属性，weightAttr[i]表示当前待编码点的已编码属性信息中第i个分量在属性距离(差异)计算中所占的权重。Among them, attrDim represents the number of attributes, for example, the number of reflectance components is 1, and the number of color components is 3. curAttr indicates the encoded attribute information of the current point to be encoded, neighborAttr indicates the encoded first attribute of the encoded candidate point, weightAttr[i] indicates the attribute distance (difference) of the i-th component in the encoded attribute information of the current point to be encoded The weight to account for in the calculation.

举例说明，以下面的例子说明步骤S200的整体过程：For example, the overall process of step S200 is illustrated with the following example:

通过几何距离的筛选得到10个已编码候选点，这10个已编码候选点都至少包括了已经编码的颜色属性和折射率属性，而待编码点也至少包括了颜色属性和折射率属性，其中待编码点的颜色属性已经被编码了，而折射率没有被编码。现在开始计算这10个已编码候选点与与待编码点之间的属性距离，计算每个已编码候选点的颜色属性与待编码点的颜色属性之间的差值，以此得到已编码候选点与待编码点之间的属性距离。Through the screening of geometric distance, 10 coded candidate points are obtained, and these 10 coded candidate points all include at least the coded color attribute and refractive index attribute, and the points to be coded also include at least the color attribute and refractive index attribute, where The color attribute of the point to be encoded has been encoded, but the refractive index has not been encoded. Now start to calculate the attribute distance between these 10 encoded candidate points and the point to be encoded, and calculate the difference between the color attribute of each encoded candidate point and the color attribute of the point to be encoded, so as to obtain the encoded candidate The attribute distance between the point and the point to be encoded.

S300，依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点。S300. According to the geometric distance and the attribute distance, obtain an encoded matching point among the encoded candidate points that matches the point to be encoded.

无论是几何距离还是属性距离，通过这两个因素所找到的点都是与待编码点有一定相关性的点。步骤S300包括如下的步骤S301和S302：Regardless of the geometric distance or the attribute distance, the points found through these two factors are all points that have a certain correlation with the points to be encoded. Step S300 includes the following steps S301 and S302:

S301，对所述待编码点与所述已编码点中的所述已编码候选点之间的所述几何距离以及所述待编码点与所述已编码候选点之间的属性距离进行加权计算，得到所述待编码点和所述已编码候选点之间的综合距离。S301. Perform weighted calculation on the geometric distance between the point to be encoded and the encoded candidate point in the encoded point and the attribute distance between the point to be encoded and the encoded candidate point , to obtain the integrated distance between the point to be encoded and the encoded candidate point.

本实施例利用综合理解去衡量已编码候选点与待编码点之间的相关性，综合距离越小，两点的相关性越强，计算综合距离的过程如下：This embodiment uses comprehensive understanding to measure the correlation between the encoded candidate point and the point to be encoded. The smaller the integrated distance, the stronger the correlation between the two points. The process of calculating the integrated distance is as follows:

其中，dis表示综合距离，disGeom表示几何距离，disAttr[i]表示已编码候选点与待编码点共同的第i种已编码属性信息距离。codedAttrNum表示已编码属性信息所包含的属性信息个数。maxGeom表示几何信息的最大值，maxAttr[i]表示第i种已编码属性信息的最大值。lGeom表示几何距离在综合距离计算中的权重，lAttr[i]表示第i个已编码的属性信息的属性息距离在综合距离计算中的权重。(lGeom>＝0，lAttr[i]>＝0)。本方案中，权重lAttr[i]的值可以按照如下公式进行计算：Among them, dis represents the comprehensive distance, disGeom represents the geometric distance, and disAttr[i] represents the i-th coded attribute information distance between the coded candidate point and the point to be coded. codedAttrNum indicates the number of attribute information included in the encoded attribute information. maxGeom represents the maximum value of geometric information, and maxAttr[i] represents the maximum value of the i-th encoded attribute information. lGeom represents the weight of the geometric distance in the comprehensive distance calculation, and lAttr[i] represents the weight of the attribute information distance of the i-th encoded attribute information in the comprehensive distance calculation. (lGeom>=0, lAttr[i]>=0). In this scheme, the value of weight lAttr[i] can be calculated according to the following formula:

lAttr[i]＝a·attrQuantParam+blAttr[i]=a·attrQuantParam+b

其中a、b为任意实数，attrQuantParam表示属性量化参数。Among them, a and b are any real numbers, and attrQuantParam represents the attribute quantization parameter.

S302，依据所述综合距离，得到所述已编码匹配点。S302. Obtain the encoded matching point according to the integrated distance.

步骤S301计算出已编码候选点与待编码点之间的综合距离之后，再从已编码候选点挑选出与待编码点相关程度更大的已编码匹配点。In step S301, after calculating the comprehensive distance between the encoded candidate points and the points to be encoded, the encoded matching points that are more closely related to the points to be encoded are selected from the encoded candidate points.

挑选出已编码匹配点相当于更新待编码点在点云中的最近邻，更新最近邻的过程如下：Selecting the encoded matching point is equivalent to updating the nearest neighbor of the point to be encoded in the point cloud. The process of updating the nearest neighbor is as follows:

在计算得到当前待编码点与邻居已编码候选点的综合距离后，更新最近邻列表，即将计算得到的综合距离和最近邻中点的综合距离进行比较，若小于其中任意一个最近邻的综合距离，则将该最近邻点替换为当前邻居候选点。After calculating the comprehensive distance between the current point to be coded and the neighbor coded candidate point, update the nearest neighbor list, that is, compare the calculated comprehensive distance with the comprehensive distance of the nearest neighbor midpoint, if it is less than the comprehensive distance of any one of the nearest neighbors , then replace the nearest neighbor point with the current neighbor candidate point.

当步骤S200中的属性距离为属性之间的差异时，可以将综合距离小于设定值所对应的已编码候选点标记为已编码匹配点。When the attribute distance in step S200 is the difference between attributes, the encoded candidate points whose integrated distance is smaller than the set value may be marked as encoded matching points.

S400，依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。S400. Predict the attribute to be encoded corresponding to the point to be encoded according to the encoded matching point, and obtain a predicted value.

已编码匹配点是与待编码点相关性比较大的已编码点，通过已编码匹配点的属性信息可以对待编码点的待编码属性信息进行预测。步骤S400包括如下的步骤S401、S402：The coded matching point is a coded point that has a relatively high correlation with the point to be coded, and the attribute information of the coded point to be coded can be predicted through the attribute information of the coded matching point. Step S400 includes the following steps S401 and S402:

S401，依据所述已编码匹配点所对应的已编码属性信息，得到所述已编码匹配点所对应的已编码属性信息中的已编码第二属性，所述已编码第二属性与所述待编码属性所匹配。具体过程为：依据所述待编码点与所述已编码匹配点之间的综合距离和所述已编码第二属性，对所述待编码点所对应的待编码属性进行预测，也就是将已编码匹配点与待编码点之间的综合距离作为已编码第二属性的权重，然后进行加权计算。S401. According to the encoded attribute information corresponding to the encoded matching point, obtain the encoded second attribute in the encoded attribute information corresponding to the encoded matching point, and the encoded second attribute is related to the waiting Encoding attributes are matched. The specific process is: according to the comprehensive distance between the point to be encoded and the encoded matching point and the encoded second attribute, predict the attribute to be encoded corresponding to the point to be encoded, that is, the The comprehensive distance between the encoded matching point and the point to be encoded is used as the weight of the encoded second attribute, and then weighted calculation is performed.

计算权重的具体过程如下：The specific process of calculating the weight is as follows:

w[i]第i个已编码匹配点(已编码匹配点不只一个，需要计算每个的权重)在待编码点的待编码属性编码过程中的权重，lGeom2表示几何距离在待编码属性编码过程中的权重(lGeom2>0)，其可以等于S301中计算综合距离的lGeom。lAttr2[i][j]表示第i个已编码匹配点的第j个已编码属性信息的属性距离在属性预测值计算中的权重(lAttr2[j]>＝0),其可以等于S301中计算综合距离的lAttr[j]。maxGeom表示几何信息的最大值，maxAttr[j]表示第j种已编码属性信息的最大值。codedAttrNum表示已编码的不同类型的属性信息个数。disGeom[i]表示第i 个最近邻点和待编码点的几何信息距离，disAttr[i][j]表示第i个已编码匹配点和待编码点的第j个已编码属性的属性信息距离。w[i] The weight of the i-th encoded matching point (there is more than one encoded matching point, and the weight of each needs to be calculated) in the encoding process of the attribute to be encoded of the point to be encoded, lGeom2 indicates the geometric distance in the encoding process of the attribute to be encoded The weight in (lGeom2>0), which may be equal to the lGeom of calculating the integrated distance in S301. lAttr2[i][j] represents the weight of the attribute distance of the j-th encoded attribute information of the i-th encoded matching point in the calculation of the attribute prediction value (lAttr2[j]>=0), which can be equal to the calculation in S301 lAttr[j] of the combined distance. maxGeom represents the maximum value of geometric information, and maxAttr[j] represents the maximum value of jth encoded attribute information. codedAttrNum indicates the number of encoded attribute information of different types. disGeom[i] represents the geometric information distance between the i-th nearest neighbor point and the point to be encoded, and disAttr[i][j] represents the attribute information distance between the i-th encoded matching point and the j-th encoded attribute of the point to be encoded .

S402，依据所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测。S402. Predict the second attribute to be encoded corresponding to the point to be encoded according to the encoded second attribute corresponding to the encoded matching point.

具体预测过程如下：The specific prediction process is as follows:

计算得到每一个已编码匹配点的权重之后，将每一个已编码匹配点按照权重进行加权求和得到当前待编码点的属性预测值(后续再根据预测值对待编码点的待编码属性进行编码)。其中属性预测值的计算公式如下所示：After the weight of each encoded matching point is calculated, each encoded matching point is weighted and summed according to the weight to obtain the predicted value of the attribute of the current point to be encoded (the attribute to be encoded of the point to be encoded is then encoded according to the predicted value) . The calculation formula of the attribute prediction value is as follows:

其中，w[i]为第i个已编码匹配点的权重。predAttr表示当前待编码点的属性预测值，nearestNeighNum表示最近邻点个数，reconAttr[i]表示第i个最近邻点的重建属性值。Among them, w[i] is the weight of the i-th encoded matching point. predAttr indicates the attribute prediction value of the current point to be encoded, nearestNeighNum indicates the number of nearest neighbor points, and reconAttr[i] indicates the reconstructed attribute value of the i-th nearest neighbor point.

举例说明，以下面的例子说明步骤S400的整体过程：For example, the overall process of step S400 is illustrated with the following example:

有3个已编码匹配点，已编码匹配点包括颜色属性和折射率属性，计算3个已编码匹配点与待编码点之间的综合距离，综合距离就是已编码匹配点与待编码点之间的几何距离以及已编码匹配点的折射率属性与待编码的已编码的折射率属性之间的属性距离这两个距离的加权，得到综合距离之后，开始计算待编码点的待编码属性(颜色属性)的预测值，之后再根据预测值进行编码，其中计算预测值是通过综合距离计算各个已编码匹配点所对应的颜色属性的权重，之后再进行相加计算，就得到待编码点的待编码属性(颜色属性)的预测值。本实施例采用了已编码的颜色属性和折射率属性以及待编码点的已编码折射率属性，得到待编码的待编码颜色属性的预测值，从而实现跨属性类别的属性值预测。There are 3 encoded matching points. The encoded matching points include color attributes and refractive index attributes. Calculate the comprehensive distance between the 3 encoded matching points and the points to be encoded. The integrated distance is the distance between the encoded matching points and the points to be encoded. The geometric distance of the coded matching point and the attribute distance between the coded refractive index attribute and the coded refractive index attribute to be coded are weighted. After the comprehensive distance is obtained, the coded attribute of the coded point (color attribute), and then encode according to the predicted value, wherein the calculation of the predicted value is to calculate the weight of the color attribute corresponding to each coded matching point through the comprehensive distance, and then perform the addition calculation to obtain the to-be-coded point. The predicted value of the encoded attribute (color attribute). In this embodiment, the coded color attribute and refractive index attribute and the coded refractive index attribute of the point to be coded are used to obtain the predicted value of the coded color attribute to be coded, so as to realize the property value prediction across attribute categories.

当本实施例中，当用反射率去计算综合距离，再依据综合距离预测颜色属性时，lGeom取值为1，lAttr的值通过属性量化参数(attrQuantParam)进行计算，When in the present embodiment, when using the reflectivity to calculate the comprehensive distance, and then predicting the color attribute according to the comprehensive distance, the value of lGeom is 1, and the value of lAttr is calculated by the attribute quantization parameter (attrQuantParam),

lAttr[i]＝a·attrQuantParam+blAttr[i]=a·attrQuantParam+b

其中参数a取值为-1/800，参数b取值为13/100。在属性预测值生成过程中，lGeom2取值为1，lAttr2取值为0，得到如图4所示的结果。The value of parameter a is -1/800, and the value of parameter b is 13/100. In the process of attribute prediction value generation, the value of lGeom2 is 1, and the value of lAttr2 is 0, and the result shown in Figure 4 is obtained.

当本实施例中，当用颜色属性去计算综合距离，再依据综合距离预测折射率时，几何距离和属性距离均采用曼哈顿距离。综合距离计算中，lGeom取值为1，lAttr的值通过属性量化参数(attrQuantParam)进行计算，In this embodiment, when the color attribute is used to calculate the comprehensive distance, and then the refractive index is predicted according to the comprehensive distance, both the geometric distance and the attribute distance use the Manhattan distance. In the comprehensive distance calculation, the value of lGeom is 1, and the value of lAttr is calculated by the attribute quantization parameter (attrQuantParam).

lAttr[i]＝a·attrQuantParam+blAttr[i]=a·attrQuantParam+b

其中参数a取值为-1/800，参数b取值为8/100。在属性预测值生成过程中，lGeom2取值为1，lAttr2取值为0，得到如图5所示的结果。The value of parameter a is -1/800, and the value of parameter b is 8/100. In the process of attribute prediction value generation, the value of lGeom2 is 1, and the value of lAttr2 is 0, and the result shown in Figure 5 is obtained.

下面说明一下本发明的点云预测的整体过程：The overall process of the point cloud prediction of the present invention is described below:

如图3所示，首先判断是否遍历完点云中的所有邻居候选点(已编码点)，所谓的遍历完就是是否已经计算出已编码点与待编码点之间的综合距离，如果已经遍历完，则直接通过已编码点计算待编码点的待编码属性的预测值。如果没有遍历完，则计算已编码点与待编码点之间的几何距离和属性距离，再结合几何距离和属性距离计算综合距离，之后再根据综合距离判断哪些已编码点是待编码点的最近点(已编码匹配点)，将已编码匹配点作为待编码点的最近邻。As shown in Figure 3, it is first judged whether all neighbor candidate points (encoded points) in the point cloud have been traversed. After that, the predicted value of the attribute to be encoded of the point to be encoded is directly calculated through the encoded point. If the traversal is not complete, calculate the geometric distance and attribute distance between the encoded point and the point to be encoded, and then combine the geometric distance and attribute distance to calculate the comprehensive distance, and then judge which encoded points are the closest to the point to be encoded according to the comprehensive distance point (encoded matching point), the encoded matching point is used as the nearest neighbor of the point to be encoded.

本实施例还提供一种点云编码方法，包括：This embodiment also provides a point cloud encoding method, including:

获取待编码点；Obtain the points to be encoded;

本实施例还提供一种点云解码方法，包括：This embodiment also provides a point cloud decoding method, including:

依据上述所述的点云预测方法，得到所述待解码点所对应的预测值；According to the point cloud prediction method described above, the predicted value corresponding to the point to be decoded is obtained;

依据所述码流，得到带解码点对应的预测残差值；Obtaining a prediction residual value corresponding to a decoding point according to the code stream;

综上，本发明在几何距离的基础上又增加了属性距离(属性之间的差异)，之后将几何距离和属性距离进行结合，得到点云中与待编码点所匹配的已编码点，只有找到合适的已编码点才能对待编码点进行准确的编码。本发明的属性距离是待编码点所对应的已编码属性与已编码点所对应的已编码属性之间的距离，这两个已编码属性之间存在相关性，而待编码点所对应的已编码属性和待编码点所对应的待编码属性又不同，从而实现了跨属性编码。本发明采用跨属性编码充分考虑了已编码点属性与待编码属性之间的联系，从而在基于几何距离的基础上结合了属性距离以降低待编码点与已编码点之间的属性残差值，进而提高点云编码质量和效率。In summary, the present invention adds the attribute distance (the difference between attributes) on the basis of the geometric distance, and then combines the geometric distance and the attribute distance to obtain the encoded point in the point cloud that matches the point to be encoded. Only Only by finding the appropriate coded point can the coded point to be coded be accurately coded. The attribute distance in the present invention is the distance between the encoded attribute corresponding to the point to be encoded and the encoded attribute corresponding to the encoded point. The attributes to be encoded corresponding to the encoding attributes and the points to be encoded are different, so that cross-attribute encoding is realized. The present invention uses cross-attribute coding to fully consider the connection between the encoded point attribute and the attribute to be encoded, so that the attribute distance is combined on the basis of the geometric distance to reduce the residual value of the attribute between the point to be encoded and the encoded point , thus improving the quality and efficiency of point cloud encoding.

示例性装置Exemplary device

本实施例还提供一种点云预测装置，所述装置包括如下组成部分：This embodiment also provides a point cloud prediction device, which includes the following components:

基于上述实施例，本发明还提供了一种终端设备，所述终端设备包括存储器、处理器及存储在所述存储器中并可在所述处理器上运行的点云预测程序，所述处理器执行所述点云预测程序时，实现上述所述的点云预测方法的步骤。Based on the above embodiments, the present invention also provides a terminal device, the terminal device includes a memory, a processor, and a point cloud prediction program stored in the memory and operable on the processor, the processor When the point cloud prediction program is executed, the steps of the above-mentioned point cloud prediction method are realized.

本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本发明所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM (EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware, and the computer programs can be stored in a non-volatile computer-readable memory In the medium, when the computer program is executed, it may include the processes of the embodiments of the above-mentioned methods. Wherein, any reference to memory, storage, database or other media used in the various embodiments provided by the present invention may include non-volatile and/or volatile memory. Nonvolatile memory can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

综上，本发明公开了点云预测、点云编码、点云解码方法及设备，所述方法包括：依据点云中的待编码点在点云中的位置与已编码点在点云中的位置，得到所述待编码点与所述已编码点之间的几何距离；依据所述已编码点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码点之间的属性距离，所述已编码点所对应的已编码属性信息与所述待编码点所对应的已编码属性信息所匹配；依据所述几何距离和所述属性距离，得到所述已编码点中与所述待编码点所匹配的已编码匹配点；依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。本发明在几何距离的基础上又增加了属性距离，之后将几何距离和属性距离进行结合，得到点云中与待编码点所匹配的已编码点，只有找到合适的已编码点才能对待编码点进行准确的编码。本发明的属性距离是待编码点所对应的已编码属性与已编码点所对应的已编码属性之间的距离，这两个已编码属性之间存在相关性，而待编码点所对应的已编码属性和待编码点所对应的待编码属性又不同，从而实现了跨属性编码。To sum up, the present invention discloses point cloud prediction, point cloud encoding, and point cloud decoding methods and devices. position to obtain the geometric distance between the point to be coded and the coded point; according to the coded attribute information corresponding to the coded point and the coded attribute information corresponding to the coded point, the The attribute distance between the point to be encoded and the encoded point, the encoded attribute information corresponding to the encoded point matches the encoded attribute information corresponding to the encoded point; according to the geometric distance and the The attribute distance is used to obtain the encoded matching point that matches the to-be-encoded point among the encoded points; according to the encoded matching point, the to-be-encoded attribute corresponding to the to-be-encoded point is predicted to obtain the predicted value. The present invention increases the attribute distance on the basis of the geometric distance, and then combines the geometric distance and the attribute distance to obtain the coded point that matches the point to be coded in the point cloud. Only when a suitable coded point is found can the coded point be treated for accurate coding. The attribute distance in the present invention is the distance between the encoded attribute corresponding to the point to be encoded and the encoded attribute corresponding to the encoded point. The attributes to be encoded corresponding to the encoding attributes and the points to be encoded are different, so that cross-attribute encoding is realized.

最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

一种点云预测方法，其特征在于，包括：A point cloud prediction method, characterized in that, comprising:

根据点云中的已编码点，得到所述已编码点中的已编码候选点；Obtaining encoded candidate points in the encoded points according to the encoded points in the point cloud;

依据点云中的待编码点在点云中的位置与所述已编码候选点在点云中的位置，得到所述待编码点与所述已编码候选点之间的几何距离；According to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud, the geometric distance between the point to be encoded and the encoded candidate point is obtained;

依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离；Obtaining an attribute distance between the point to be encoded and the candidate point to be encoded according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the to-be-encoded point;

依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点；Obtaining an encoded matching point among the encoded candidate points that matches the point to be encoded according to the geometric distance and the attribute distance;

依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。According to the encoded matching point, predict the attribute to be encoded corresponding to the point to be encoded to obtain a predicted value.
如权利要求1所述的点云预测方法，其特征在于，所述根据点云中的已编码点，得到所述已编码点中的已编码候选点，包括方法一或方法二或方法三：The point cloud prediction method according to claim 1, wherein, according to the encoded points in the point cloud, the encoded candidate points in the encoded points are obtained, including method one or method two or method three:

方法一：将所有所述已编码点作为所述已编码候选点；Method 1: using all the encoded points as the encoded candidate points;

方法二：将所有所述已编码点按照对应的希尔伯特码排序，根据所述希尔伯特码排序选择得到所述已编码候选点；Method 2: sorting all the encoded points according to the corresponding Hilbert codes, and selecting the encoded candidate points according to the sorting of the Hilbert codes;

方法三：将所有所述已编码点按照对应的莫顿码排序，根据所述莫顿码排序选择得到所述已编码候选点。Method 3: sort all the encoded points according to the corresponding Morton codes, and select the encoded candidate points according to the sorting of the Morton codes.
如权利要求1所述的点云预测方法，其特征在于：所述依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离，包括：The point cloud prediction method according to claim 1, characterized in that: according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the point to be encoded, the to-be-encoded point is obtained. The attribute distance between the coded point and the coded candidate point, including:

依据所述已编码点中的所述已编码候选点所对应的已编码属性信息，得到所述已编码候选点所对应的已编码属性信息中的已编码第一属性；Obtaining the encoded first attribute in the encoded attribute information corresponding to the encoded candidate point according to the encoded attribute information corresponding to the encoded candidate point in the encoded point;

依据所述待编码点所对应的所述已编码属性信息，得到所述待编码点所对应的所述已编码属性信息中的所述待编码点所对应的已编码第一属性；According to the encoded attribute information corresponding to the to-be-encoded point, obtain the encoded first attribute corresponding to the to-be-encoded point in the encoded attribute information corresponding to the to-be-encoded point;

依据所述已编码候选点所对应的已编码第一属性和所述待编码点所对应的已编码第一属性，得到所述待编码点与所述已编码候选点之间的差值的函数值作为属性距离。According to the encoded first attribute corresponding to the encoded candidate point and the encoded first attribute corresponding to the to-be-encoded point, the function of obtaining the difference between the to-be-encoded point and the encoded candidate point Value as attribute distance.
如权利要求1所述的点云预测方法，其特征在于，所述依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点，包括：The point cloud prediction method according to claim 1, wherein, according to the geometric distance and the attribute distance, the encoded matching points that match the points to be encoded among the encoded candidate points are obtained ,include:

对所述待编码点与所述已编码点中的所述已编码候选点之间的所述几何距离以及所述待编码点与所述已编码候选点之间的属性距离进行加权计算，得到所述待编码点和所述已编码候选点之间的综合距离；performing weighted calculation on the geometric distance between the point to be coded and the coded candidate point in the coded point and the attribute distance between the point to be coded and the coded candidate point to obtain The integrated distance between the points to be encoded and the encoded candidate points;

依据所述综合距离，按照距离大小得到所述已编码匹配点。According to the comprehensive distance, the coded matching points are obtained according to the distance.
如权利要求1所述的点云预测方法，其特征在于，所述依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值，包括：The point cloud prediction method according to claim 1, wherein, according to the encoded matching point, predicting the attribute to be encoded corresponding to the point to be encoded to obtain a predicted value includes:

依据所述已编码匹配点所对应的已编码属性信息，得到所述已编码匹配点所对应的已编码属性信息中的已编码第二属性；Obtaining the encoded second attribute in the encoded attribute information corresponding to the encoded matching point according to the encoded attribute information corresponding to the encoded matching point;

依据所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Predict the second attribute to be encoded corresponding to the point to be encoded according to the encoded second attribute corresponding to the encoded matching point to obtain a prediction value.
如权利要求5所述的点云预测方法，其特征在于，所述依据所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行进行预测，得到预测值，包括方法一或方法二：The point cloud prediction method according to claim 5, wherein, according to the encoded second attribute corresponding to the encoded matching point, the second attribute to be encoded corresponding to the point to be encoded is Make predictions and get predicted values, including method 1 or method 2:

方法一：依据所述待编码点与所述已编码匹配点之间的几何距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值；Method 1: According to the geometric distance between the point to be encoded and the encoded matching point and the encoded second attribute corresponding to the encoded matching point, the to-be-encoded point corresponding to the to-be-encoded point is The second attribute is predicted and the predicted value is obtained;

方法二：依据所述待编码点与所述已编码匹配点之间的综合距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Method 2: According to the comprehensive distance between the point to be coded and the coded matching point and the coded second attribute corresponding to the coded matching point, the to-be-coded point corresponding to the coded point The second attribute is predicted and the predicted value is obtained.
如权利要求6所述的点云预测方法，其特征在于，所述依据所述待编码点与所述已编码匹配点之间的综合距离以及所述已编码匹配点所对应的所述已编码第二属性，对所述待编码点所对应的待编码第二属性进行预测，得到预测值，包括：The point cloud prediction method according to claim 6, characterized in that, according to the comprehensive distance between the point to be encoded and the encoded matching point and the encoded matching point corresponding to the encoded The second attribute is to predict the second attribute to be encoded corresponding to the point to be encoded to obtain a predicted value, including:

依据所述综合距离，得到所述已编码第二属性所对应的权重；Obtaining the weight corresponding to the encoded second attribute according to the comprehensive distance;

依据所述已编码第二属性所对应的权重和所述已编码第二属性所对应的属性值，对所述待编码点所对应的待编码第二属性进行预测，得到预测值。Predict the second attribute to be encoded corresponding to the point to be encoded according to the weight corresponding to the encoded second attribute and the attribute value corresponding to the encoded second attribute, to obtain a predicted value.
一种点云编码方法，其特征在于，包括：A point cloud coding method is characterized in that, comprising:

获取待编码点；Obtain the points to be encoded;

依据权利要求1-7任一项所述的点云预测方法，得到所述待编码点所对应的预测值；According to the point cloud prediction method described in any one of claims 1-7, the predicted value corresponding to the point to be encoded is obtained;

获取所述待编码点所对应的待编码第二属性的设定值；Acquiring the set value of the second attribute to be encoded corresponding to the point to be encoded;

依据所述待编码第二属性的设定值和所述预测值，得到所述设定值与所述预测值之间的预测残差值；Obtaining a prediction residual value between the set value and the predicted value according to the set value of the second attribute to be encoded and the predicted value;

依据所述预测残差值，对所述待编码点所对应的待编码第二属性进行编码，得到码流。According to the prediction residual value, encode the second attribute to be encoded corresponding to the point to be encoded to obtain a code stream.
一种点云解码方法，其特征在于，包括：A point cloud decoding method is characterized in that, comprising:

获取待解码点所对应的码流；Obtain the code stream corresponding to the point to be decoded;

依据权利要求1-7任一项所述的点云预测方法，得到所述待解码点所对应的预测值；According to the point cloud prediction method described in any one of claims 1-7, the predicted value corresponding to the point to be decoded is obtained;

依据所述码流，得到所述待解码点对应的预测残差值；Obtaining a prediction residual value corresponding to the point to be decoded according to the code stream;

依据所述预测残差值和所述预测值，得到所述待解码点的第二属性重建值。Obtain a second attribute reconstruction value of the point to be decoded according to the prediction residual value and the prediction value.
一种点云预测装置，其特征在于，所述装置包括如下组成部分：A point cloud prediction device, characterized in that the device comprises the following components:

已编码候选点确定模块，用于根据点云中的已编码点，得到所述已编码点中的已编码候选点；The coded candidate point determination module is used to obtain the coded candidate points in the coded points according to the coded points in the point cloud;

几何距离计算模块，用于依据点云中的待编码点在点云中的位置与所述已编码候选点在点云中的位置，得到所述待编码点与所述已编码候选点之间的几何距离；The geometric distance calculation module is used to obtain the distance between the point to be encoded and the encoded candidate point according to the position of the point to be encoded in the point cloud and the position of the encoded candidate point in the point cloud. the geometric distance;

属性距离计算模块，用于依据所述已编码候选点所对应的已编码属性信息和所述待编码点所对应的已编码属性信息，得到所述待编码点与所述已编码候选点之间的属性距离；An attribute distance calculation module, configured to obtain the distance between the point to be encoded and the candidate point to be encoded according to the encoded attribute information corresponding to the encoded candidate point and the encoded attribute information corresponding to the point to be encoded. attribute distance;

匹配点计算模型，用于依据所述几何距离和所述属性距离，得到所述已编码候选点中与所述待编码点所匹配的已编码匹配点；A matching point calculation model, used to obtain an encoded matching point that matches the point to be encoded among the encoded candidate points according to the geometric distance and the attribute distance;

预测模块，用于依据所述已编码匹配点，对所述待编码点所对应的待编码属性进行预测，得到预测值。The prediction module is configured to predict the attribute to be encoded corresponding to the point to be encoded according to the encoded matching point, and obtain a predicted value.
一种终端设备，其特征在于，所述终端设备包括存储器、处理器及存储在所述存储器中并可在所述处理器上运行的点云预测程序，所述处理器执行所述点云预测程序时，实现如权利要求1-7任一项所述的点云预测方法的步骤。A terminal device, characterized in that the terminal device includes a memory, a processor, and a point cloud prediction program stored in the memory and operable on the processor, and the processor executes the point cloud prediction program, realize the steps of the point cloud prediction method as described in any one of claims 1-7.
一种计算机可读存储介质，其特征在于，所述计算机可读存储介质上存储有点云预测程序，所述点云预测程序被处理器执行时，实现如权利要求1-7任一项所述的点云预测方法的步骤。A computer-readable storage medium, characterized in that a point cloud prediction program is stored on the computer-readable storage medium, and when the point cloud prediction program is executed by a processor, it realizes any one of claims 1-7 The steps of the point cloud prediction method.