CN117319683A - Point cloud attribute compression method based on texture-driven graph sparsity optimization - Google Patents
Point cloud attribute compression method based on texture-driven graph sparsity optimization Download PDFInfo
- Publication number
- CN117319683A CN117319683A CN202311501797.9A CN202311501797A CN117319683A CN 117319683 A CN117319683 A CN 117319683A CN 202311501797 A CN202311501797 A CN 202311501797A CN 117319683 A CN117319683 A CN 117319683A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- texture
- graph
- optimal
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 36
- 230000006835 compression Effects 0.000 title claims abstract description 35
- 238000007906 compression Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 40
- 230000009466 transformation Effects 0.000 claims abstract description 18
- 238000005516 engineering process Methods 0.000 claims abstract description 7
- 238000010586 diagram Methods 0.000 claims abstract description 4
- 238000013139 quantization Methods 0.000 claims abstract description 4
- 238000001228 spectrum Methods 0.000 claims abstract description 4
- 230000001131 transforming effect Effects 0.000 claims abstract description 4
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform 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
- H04N19/91—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
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention provides a graph sparsity optimization method based on texture driving, which comprises the following steps: s1, dividing blocks of a point cloud; s2, constructing a plurality of nearest neighbor matrixes for each block to serve as candidate adjacent matrixes; s3, counting the texture complexity of each candidate adjacent matrix and determining an optimal adjacent matrix according to an energy minimization criterion; s4, determining a weight coefficient according to the mathematical relationship between the adjacent matrix and the weight coefficient which are already deduced; s5, bringing the adjacent matrix and the weight coefficient into a code rate loss optimization function to obtain an optimal weight matrix, and taking the optimal weight matrix as an optimal graph structure of the block; s6, diagram transformation: transforming the point cloud attribute values from an original threshold to a spectrum threshold based on an optimal graph structure by using a graph Fourier transform technology to obtain transformation coefficients; s7, the transformation coefficient is written into the code stream file through quantization and entropy coding. The method can be used for realizing efficient point cloud attribute compression.
Description
Technical Field
The invention relates to the field of point cloud compression, in particular to a point cloud attribute compression method based on texture-driven graph sparsity optimization.
Background
Three-dimensional point clouds are an important representation of real world digitization and have been widely used in the fields of autopilot, virtual reality, and digital museums, where point clouds are a set of points containing geometric and attribute information (e.g., color and reflectivity) that simulate the external surfaces of various scenes and objects.
The difficulty of solving the problems and the defects is as follows: with the development of point cloud acquisition equipment, the resolution of point cloud is increasing sharply, the deployment of 3D application programs is difficult to realize due to the increase of data volume, and how to efficiently compress the point cloud data has become an important problem.
Recently, various compression methods for point cloud attribute information have been proposed, and graph transformation technology is one important branch. Setting an adaptive map structure according to data characteristics is a key for obtaining efficient performance by a map transformation technology. Currently, there are two main ways of defining the graph structure, respectively: a graph structure based on a model and a graph structure based on code rate loss optimization. However, the design of these approaches neglects the influence of sparsity of the thumbnail on the encoder.
The meaning of solving the problems and the defects is as follows: the sparsity of the appropriate graph can reduce encoder time complexity while maintaining efficient compression performance. Therefore, sparsity of the graph is a key factor to further enhance the compression performance based on the graph transformation.
Disclosure of Invention
The invention provides a point cloud attribute compression method based on texture-driven graph sparsity optimization, which is used for realizing efficient point cloud attribute compression.
The technical scheme of the invention is as follows:
the invention discloses a point cloud attribute compression method based on texture-driven graph sparsity optimization, which comprises the following steps of: s1, dividing blocks of a point cloud; s2, constructing a plurality of nearest neighbor matrixes for each block to serve as candidate adjacent matrixes; s3, counting the texture complexity of each candidate adjacent matrix and determining an optimal adjacent matrix according to an energy minimization criterion; s4, determining a weight coefficient according to the mathematical relationship between the adjacent matrix and the weight coefficient which are already deduced; s5, bringing the adjacent matrix and the weight coefficient into a code rate loss optimization function to obtain an optimal weight matrix, and taking the optimal weight matrix as an optimal graph structure of the block; s6, diagram transformation: transforming the point cloud attribute values from an original threshold to a spectrum threshold based on an optimal graph structure by using a graph Fourier transform technology to obtain transformation coefficients; s7, the transformation coefficient is written into the code stream file through quantization and entropy coding.
Optionally, in the above method for compressing the point cloud attribute based on the sparseness optimization of the texture-driven graph, in step S1, an input point cloud is given, and is divided into a plurality of small blocks by using a three-dimensional KD tree structure, and each small block is used as an independent encoding unit.
Optionally, in the above-mentioned method for compressing the point cloud attribute based on the sparseness optimization of the texture-driven graph, in step S2, for each block, a connection relationship of points is established by adopting n nearest neighbors, n=1, 2, … ….
Optionally, in the above-mentioned point cloud attribute compression method based on texture-driven graph sparsity optimization, in step S3, the texture complexity is obtained by using a color difference calculation of the entire block, where the texture complexity is measured by using CIEDE2000 standard.
Optionally, in the above-mentioned point cloud attribute compression method based on texture driving graph sparsity optimization, in step S4, the weight coefficient λ is determined according to the relationship between the n nearest neighbor adjacency matrices that have been derived and the weight coefficient λ, where the relationship between the n and λ two variables is as follows:
optionally, in the above method for compressing a point cloud attribute based on texture-driven graph sparsity optimization, in step S6, a code rate loss optimization function is:
z is the n nearest neighbor adjacency matrix.
Optionally, in the above-mentioned point cloud attribute compression method based on texture driving graph sparsity optimization, in step S7, the transform coefficients are quantized into integers, and the quantized transform coefficients are entropy-encoded into a code stream file.
According to the technical scheme of the invention, the beneficial effects are that:
according to the texture-driven graph sparsity optimization-based point cloud attribute compression method, the influence of sparsity on the construction of the graph structure is considered, so that the graph structure with low complexity and high compression performance is constructed. In addition, the sparsity of the texture complexity optimization graph is utilized, and performance deviation caused by the fact that the existing scheme only utilizes the point cloud geometry is avoided.
For a better understanding and explanation of the conception, working principle and inventive effect of the present invention, the present invention is described in detail below by way of specific examples with reference to the accompanying drawings, in which:
drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
FIG. 1 is a flow chart of a point cloud attribute compression method based on texture driven graph sparsity optimization of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples, in order to make the objects, technical methods and advantages of the present invention more apparent. These examples are illustrative only and are not limiting of the invention.
In order to reduce the time complexity based on graph transformation and obtain higher compression performance, the texture-driven graph sparsity optimization-based point cloud attribute compression method is used for realizing efficient point cloud attribute compression and obtaining the current optimal compression performance.
The invention relates to a point cloud attribute compression method based on texture-driven graph sparsity optimization, which is shown in fig. 1 and comprises the following steps:
s1, block division based on a K-dimensional tree (KD-tree): and performing block division on the point cloud by using a KD-tree structure.
Given an input point cloud, firstly, the input point cloud is divided into a plurality of small blocks by adopting a KD-tree structure, and each small block is used as an independent coding unit.
In particular, the point cloud is partitioned into blocks using a three-dimensional KD-tree structure, each non-leaf node may be partitioned into two subspaces by a hyperplane, and each subspace may be recursively partitioned in the same manner. The KD-tree is segmented along the coordinate axes, and all hyperplanes are perpendicular to the corresponding coordinate axes. For example, the x-axis division is performed, and only a certain x value is needed to determine the position of the hyperplane, and the hyperplane divides the original node space into two subspaces, wherein the x value of all points in one subspace is smaller than that of all points in the other subspace. Each leaf node of the KD-tree is a block.
S2, constructing a plurality of nearest neighbor matrixes for each block to serve as candidate adjacent matrixes;
for each block, multiple nearest neighbor connection methods are adopted (a graph adjacency matrix based on 1 nearest neighbor is constructed, a graph adjacency matrix based on 2 nearest neighbors is constructed, …, and a graph adjacency matrix based on 15 nearest neighbors is constructed, as shown in fig. 1).
Specifically, for each block, the connection relation of points is established in n nearest neighbors (n=1, 2, … …). Thus, each block has a plurality of candidate adjacency matrix forms.
S3, counting the texture complexity of each candidate adjacent matrix and determining an optimal adjacent matrix according to an energy minimization criterion;
in the step, selecting a nearest neighbor connection mode which minimizes the texture complexity of the current block to construct an adjacency matrix of the graph; texture complexity is obtained by using the color difference calculation of the whole block, wherein the texture complexity is measured by using CIEDE2000 standard.
S4, determining a weight coefficient according to the mathematical relationship between the adjacent matrix and the weight coefficient which are already deduced;
the weight coefficient lambda is determined according to the relation between the n nearest neighbor adjacency matrixes and the weight coefficient lambda, wherein the relation between the n and lambda variables is as follows:
s5, bringing the adjacent matrix and the weight coefficient into a code rate loss optimization function to obtain an optimal weight matrix, and taking the optimal weight matrix as an optimal graph structure of the block;
specifically, carrying out optimization iteration by taking the adjacency matrix and the weight coefficient into a code rate loss optimization function, and obtaining an optimal graph structure after stopping the iteration; wherein, the code rate loss optimization function is:
z is the n nearest neighbor adjacency matrix.
S6, diagram transformation: transforming the point cloud attribute values from an original threshold to a spectrum threshold (from a space domain to a map domain) based on an optimal map structure by using a map Fourier transform technology to obtain transform coefficients;
s7, the transformation coefficient is written into the code stream file through quantization and entropy coding. Specifically, the transform coefficients are quantized into integers (quantized coefficients), and the quantized transform coefficients (i.e., quantized coefficients) are entropy encoded into a bitstream file.
Given an input point cloud, firstly, the input point cloud is divided into a plurality of small blocks by using a KD-tree, and each small block is used as an independent coding unit. The two parameters affecting the sparsity of each block are then determined as: binary adjacency matrix and weight matrix coefficients. For each block, a plurality of nearest neighbor connection modes (1 nearest neighbor, 2 nearest neighbors, …,15 nearest neighbors) are adopted, and then the nearest neighbor connection mode which minimizes the texture complexity of the block is selected to construct the adjacency matrix of the graph. The texture complexity is here obtained using the color difference calculation of the whole block. And then obtaining the weight coefficient according to the mathematical relationship between the derived adjacency matrix and the weight coefficient. And then, bringing the adjacent matrix and the weight coefficient into a code rate loss optimization function to obtain an optimal weight matrix. Then, the weight matrix is used as a graph structure of the block to further utilize a graph Fourier transform technology to transform the point cloud attribute values from a space domain to a map domain to obtain transform coefficients. Finally, the transform coefficients are quantized and entropy encoded and written into a bitstream file.
Effect verification
In order to verify the effect of the invention, the embodiment of the method of the invention is compared with the test result of the current point cloud attribute compression platform. The test results are shown in tables 1 and 2.
Table 1: the proposed solution compares with the compression performance of the optimal platform (PLT, RAHT, HAC)
In Table 1, PLT (G-PCCv 14): an optimal predictive transformation method; RAHT (G-PCCv 14): an optimal region adaptive hierarchical transformation; HAC: hybrid attribute encoder.
Table 2: run-time comparison (unit: s) based on graph compression algorithm
From the test results in tables 1 and 2, it can be seen that the method provided by the invention has optimal compression performance, and compared with similar compression platforms, the compression performance is averagely improved by 13.72% -23.02%. Compared with the common graph transformation mode HAC, the calculation time is only one third of that of the common graph transformation mode HAC, which proves the effectiveness of the graph sparse optimization strategy provided by the invention.
The above description is of the best mode of carrying out the inventive concept and principles of operation. The above examples should not be construed as limiting the scope of the claims, but other embodiments and combinations of implementations according to the inventive concept are within the scope of the invention.
Claims (7)
1. The texture-driven graph sparsity optimization-based point cloud attribute compression method is characterized by comprising the following steps of:
s1, dividing blocks of a point cloud;
s2, constructing a plurality of nearest neighbor matrixes for each block to serve as candidate adjacent matrixes;
s3, counting the texture complexity of each candidate adjacent matrix and determining an optimal adjacent matrix according to an energy minimization criterion;
s4, determining a weight coefficient according to the mathematical relationship between the adjacent matrix and the weight coefficient which are already deduced;
s5, bringing the adjacent matrix and the weight coefficient into a code rate loss optimization function to obtain an optimal weight matrix, and taking the optimal weight matrix as an optimal graph structure of the block;
s6, diagram transformation: transforming the point cloud attribute values from an original threshold to a spectrum threshold based on the optimal graph structure by using a graph Fourier transform technology to obtain transformation coefficients; and
s7, the transformation coefficient is written into a code stream file through quantization and entropy coding.
2. The texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein in step S1, an input point cloud is given and divided into a plurality of small blocks by a three-dimensional tree structure, and each small block serves as an independent coding unit.
3. The texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein in step S2, a connection relationship of points is established by adopting n nearest neighbors for each block, n=1, 2, … ….
4. The texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein in step S3, the texture complexity is obtained using a color difference calculation of the entire block, and wherein the texture complexity is measured using a CIEDE2000 standard.
5. The texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein in step S4, the weight coefficient λ is determined according to a relationship between the n nearest neighbor adjacency matrices that have been derived and the weight coefficient λ, wherein the relationship between the n and λ variables is as follows:
6. the texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein the code rate loss optimization function in step S6 is:
z is the n nearest neighbor adjacency matrix.
7. The texture-driven graph sparsity optimization-based point cloud attribute compression method of claim 1, wherein in step S7, the transform coefficients are quantized to integers, and the quantized transform coefficients are entropy encoded into the code stream file.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311501797.9A CN117319683A (en) | 2023-11-10 | 2023-11-10 | Point cloud attribute compression method based on texture-driven graph sparsity optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311501797.9A CN117319683A (en) | 2023-11-10 | 2023-11-10 | Point cloud attribute compression method based on texture-driven graph sparsity optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117319683A true CN117319683A (en) | 2023-12-29 |
Family
ID=89286681
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311501797.9A Pending CN117319683A (en) | 2023-11-10 | 2023-11-10 | Point cloud attribute compression method based on texture-driven graph sparsity optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117319683A (en) |
-
2023
- 2023-11-10 CN CN202311501797.9A patent/CN117319683A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108335335B (en) | A kind of point cloud genera compression method based on enhancing figure transformation | |
| US11004240B2 (en) | Hierarchical division-based point cloud attribute compression method | |
| CN110572655B (en) | Method and equipment for encoding and decoding point cloud attribute based on neighbor weight parameter selection and transmission | |
| CN110418135B (en) | Point cloud intra-frame prediction method and device based on neighbor weight optimization | |
| Genova et al. | Local deep implicit functions for 3d shape | |
| Golla et al. | Real-time point cloud compression | |
| CN109257604B (en) | Color attribute coding method based on TMC3 point cloud encoder | |
| US10939123B2 (en) | Multi-angle adaptive intra-frame prediction-based point cloud attribute compression method | |
| Shao et al. | Attribute compression of 3D point clouds using Laplacian sparsity optimized graph transform | |
| Huang et al. | Octree-Based Progressive Geometry Coding of Point Clouds. | |
| WO2019153326A1 (en) | Intra-frame prediction-based point cloud attribute compression method | |
| Zhang et al. | Point cloud attribute compression with graph transform | |
| CN113678466A (en) | Method and apparatus for predicting point cloud attribute encoding | |
| JP7330306B2 (en) | Transform method, inverse transform method, encoder, decoder and storage medium | |
| Wen et al. | Lossy geometry compression of 3d point cloud data via an adaptive octree-guided network | |
| Genova et al. | Deep structured implicit functions | |
| CN110691243A (en) | Point cloud geometric compression method based on deep convolutional network | |
| US9819964B2 (en) | Limited error raster compression | |
| JP7386337B2 (en) | Division method, encoder, decoder and computer storage medium | |
| WO2021023206A1 (en) | Point cloud attribute prediction, encoding, and decoding method and device based on neighbor weight optimization | |
| Deng et al. | Compressing explicit voxel grid representations: fast nerfs become also small | |
| US20230237705A1 (en) | Methods for level partition of point cloud, and decoder | |
| Zhang et al. | Adaptive geometry partition for point cloud compression | |
| CN113518226A (en) | G-PCC point cloud coding improvement method based on ground segmentation | |
| CN106658004A (en) | Compression method and device based on characteristics of flat region of image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |