WO-2026092488-A1 - ENCODING METHOD, DECODING METHOD, AND RELATED DEVICE

WO2026092488A1WO 2026092488 A1WO2026092488 A1WO 2026092488A1WO-2026092488-A1

Abstract

The present application relates to the field of three-dimensional data point cloud encoding and decoding, and discloses an encoding method, a decoding method, and a related device. The encoding method comprises: an encoding end constructing a transform tree structure of a current attribute of a point cloud to be encoded; acquiring a first original attribute value of a current attribute of a placeholder child node of a current node of the transform tree structure and a reference attribute value of a reference attribute; acquiring a cross-attribute prediction residual between the current attribute and the reference attribute on the basis of the first original attribute value, the reference attribute value and a scaling parameter, the scaling parameter being used for representing the correlation between the current attribute and the reference attribute; and performing transformation, quantization and encoding on the cross-attribute prediction residual to obtain a bitstream.

Inventors

ZHANG, WEI
WANG, JUNJIE
YANG, FUZHENG
LV, Zhuoyi

Assignees

维沃移动通信有限公司

Dates

Publication Date: 20260507
Application Date: 20251029
Priority Date: 20241030

Claims (20)

An encoding method, comprising: The encoding end constructs a transformation tree structure for the current attributes of the point cloud to be encoded; Obtain the first original attribute value and the reference attribute value of the current attribute of the placeholder child node of the current node in the transformed tree structure; Based on the first original attribute value, the reference attribute value, and the scaling parameter, the cross-attribute prediction residuals of the current attribute and the reference attribute are obtained; the scaling parameter is used to characterize the correlation between the current attribute and the reference attribute. The cross-attribute prediction residuals are transformed, quantized, and encoded to obtain a bitstream.
According to the method of claim 1, wherein obtaining the cross-attribute prediction residual of the current attribute and the reference attribute based on the first original attribute value, the reference attribute value, and the scaling parameter includes: If the current node makes a prediction, then obtain the first predicted value of the current attribute and the second predicted value of the reference attribute of the current node; The first prediction residual of the current attribute is obtained based on the first original attribute value and the first predicted value; and the second prediction residual of the reference attribute is obtained based on the reference attribute value and the second predicted value. The cross-attribute prediction residual is obtained based on the first prediction residual, the second prediction residual, and the scaling parameter.
According to the method of claim 1, wherein obtaining the cross-attribute prediction residual of the current attribute and the reference attribute based on the first original attribute value, the reference attribute value, and the scaling parameter includes: If the current node does not make a prediction, the cross-attribute prediction residual is obtained based on the first original attribute value, the reference attribute value, and the scaling parameter.
The method according to any one of claims 1-3, wherein nodes in each level of the transform tree structure correspond to the same scaling parameter; or Each node in the transformation tree structure corresponds to the same scaling parameter.
The method according to any one of claims 1-3, further comprising: For the current layer of the transformed tree structure, obtain the n nodes of the k-th group; k and n are integers greater than or equal to 1. Obtain the second original attribute value of the current attribute and the third original attribute value of the reference attribute of the i-th node of the current layer; where i∈kn～(k+1)n-1 and are integers; The scaling parameter is determined based on the second original attribute value and the third original attribute value of the i-th node; wherein the scaling parameter is shared by the n nodes.
The method according to any one of claims 1-3, further comprising: For the current layer in the transformation tree structure, obtain the second original attribute value of the current attribute and the third original attribute value of the reference attribute of the i-th node of the current layer; where i is less than or equal to the number of nodes in the current layer and is a positive integer; The scaling parameter is determined based on the second original attribute value and the third original attribute value of the i-th node; wherein the scaling parameter is shared by all nodes in the current layer.
According to the method of claim 5 or 6, wherein determining the scaling parameter based on the second original attribute value and the third original attribute value of the i-th node includes: Determine the product of the second original attribute value and the third original attribute value of the i-th node, and sum the products corresponding to the nodes within the range of the i-th node; Determine the sum of squares of the third original attribute values of the nodes within the range of the i-th node; The scaling parameter is obtained by comparing the sum of the products of the nodes within the range of the i-th node with the sum of squares.
The method according to any one of claims 1-3, further comprising: For the current layer of the transformed tree structure, obtain the first attribute reconstruction data of the current attribute and the second attribute reconstruction data of the reference attribute of the j-th node of the current layer; the j-th node belongs to the already encoded and reconstructed node; j is a positive integer; The scaling parameter is determined based on the first attribute reconstruction data and the second attribute reconstruction data of the j-th node.
According to the method of claim 8, wherein determining the scaling parameter based on the first attribute reconstruction data and the second attribute reconstruction data of the j-th node includes: Determine the product of the first attribute reconstruction data and the second attribute reconstruction data of the j-th node, and sum the products corresponding to the nodes within the range of the j-th node; Determine the sum of squares of the second attribute reconstructed data of the nodes within the range of the j-th node; The scaling parameter is obtained by comparing the sum of the products of the nodes within the range of the j-th node with the sum of squares.
The method according to claim 8 or 9, wherein, The j-th node belongs to the pre-encoded and reconstructed nodes of the current layer; or The j-th node belongs to the first m child nodes whose preceding order has been reconstructed; or The j-th node belongs to the first k nodes of the current layer that have been encoded and reconstructed in the preceding sequence; m and k are both positive integers.
The method according to claim 10, wherein, The number of child nodes of the previously encoded and reconstructed nodes in the current layer is greater than or equal to q; or The number of child nodes of the first k nodes is greater than or equal to q; Where q∈[0,7] and is an integer.
The method according to claim 10, further comprising: The m or the k is passed into the bitstream.
The method according to claim 12, further comprising: The q is passed into the bitstream.
The method according to any one of claims 1-13, further comprising: The scaling parameter is passed into the bitstream.
The method according to any one of claims 1-14, wherein the current node belongs to the nodes of the first A layers of the transformation tree structure, where A is a positive integer.
The method according to claim 15, further comprising: The first parameter is passed into the bitstream; the first parameter is used to instruct the nodes of the first A layer to perform cross-attribute prediction.
A decoding method, comprising: The decoding end decodes and dequantizes the bitstream to obtain the reconstructed value of the first transform coefficient residual of the point cloud to be decoded; Construct a transformation tree structure of the current attributes of the point cloud to be decoded; Obtain the reference attribute value of the placeholder child node of the current node in the transformed tree structure; Cross-attribute prediction is performed based on the reference attribute value and scaling parameters to obtain the first attribute prediction value; the scaling parameters are used to characterize the correlation between the current attribute and the reference attribute. The attribute reconstruction values of the placeholder child nodes of the current node are obtained by performing an inverse transformation based on the first attribute prediction value and the reconstructed value of the first transformation coefficient residual.
According to the method of claim 17, the step of obtaining the first attribute prediction value based on the reference attribute value and the scaling parameter includes: If the current node allows prediction, then obtain the second predicted value of the reference attribute of the current node; Based on the second predicted value and the reference attribute value, obtain the second predicted residual of the reference attribute; The predicted value of the first attribute is obtained based on the second prediction residual and the scaling parameter.
According to the method of claim 17, the step of obtaining the first attribute prediction value based on the reference attribute value and the scaling parameter includes: If the current node does not make a prediction, then the predicted value of the first attribute is obtained based on the reference attribute value and the scaling parameter.
The method according to any one of claims 17-19, wherein, Each node in the transformation tree structure corresponds to the same scaling parameter; or The nodes in the current layer share the scaling parameters; or The scaling parameter is shared by every n nodes in the current layer, where n is an integer greater than or equal to 1.

Description

Encoding and decoding methods and related equipment Cross-reference of related applications This application claims priority to Chinese Patent Application No. 202411532908.7, filed on October 30, 2024, entitled "Encoding/Decoding Method and Related Equipment", the entire contents of which are incorporated herein by reference. Technical Field This application belongs to the field of encoding and decoding technology, specifically relating to an encoding and decoding method and related equipment. Background Technology With the continuous development of point cloud technology, the compression and encoding of point cloud data has become an important research issue. Currently, both the Audio Video Coding Standard Workgroup of China (AVS) and the Moving Picture Experts Group (MPEG) of the International Organization for Standardization are developing standards for point cloud encoding, such as Geometry-based Point Cloud Compression (G-PCC). How to further improve the performance of point cloud encoding and decoding is an urgent problem to be solved. Summary of the Invention This application provides an encoding/decoding method and related equipment that can solve the problem of low encoding efficiency in point cloud attribute encoding. Firstly, an encoding method is provided, executed by the encoding end, the method comprising: The encoding end constructs a transformation tree structure for the current attributes of the point cloud to be encoded; Obtain the first original attribute value and the reference attribute value of the current attribute of the placeholder child node of the current node in the transformed tree structure; Based on the first original attribute value, the reference attribute value, and the scaling parameter, the cross-attribute prediction residuals of the current attribute and the reference attribute are obtained; the scaling parameter is used to characterize the correlation between the current attribute and the reference attribute. The cross-attribute prediction residuals are transformed, quantized, and encoded to obtain a bitstream. Secondly, a decoding method is provided, executed by the decoding end, the method comprising: The decoding end decodes and dequantizes the bitstream to obtain the reconstructed value of the first transform coefficient residual of the point cloud to be decoded; Construct a transformation tree structure of the current attributes of the point cloud to be decoded; Obtain the reference attribute value of the placeholder child node of the current node in the transformed tree structure; Cross-attribute prediction is performed based on the reference attribute value and scaling parameters to obtain the first attribute prediction value; the scaling parameters are used to characterize the correlation between the current attribute and the reference attribute. The attribute reconstruction values of the placeholder child nodes of the current node are obtained by performing an inverse transformation based on the first attribute prediction value and the reconstructed value of the first transformation coefficient residual. Thirdly, an encoding apparatus is provided, comprising: The building module is used by the encoding end to construct the transformation tree structure of the current attributes of the point cloud to be encoded; The acquisition module is used to acquire the first original attribute value and the reference attribute value of the current attribute of the placeholder child node of the current node of the transformed tree structure; The prediction module is used to obtain the cross-attribute prediction residuals of the current attribute and the reference attribute based on the first original attribute value, the reference attribute value, and the scaling parameter; the scaling parameter is used to characterize the correlation between the current attribute and the reference attribute. The encoding module is used to transform, quantize, and encode the cross-attribute prediction residuals to obtain a bitstream. Fourthly, a decoding apparatus is provided, comprising: The parsing module is used by the decoding end to decode and dequantize the bitstream to obtain the reconstructed value of the first transform coefficient residual of the point cloud to be decoded; A construction module is used to construct the transformation tree structure of the current attributes of the point cloud to be decoded; The acquisition module is used to acquire the reference attribute values of the placeholder child nodes of the current node in the transformed tree structure. The prediction module is used to perform cross-attribute prediction based on the reference attribute value and the scaling parameter to obtain a first attribute prediction value; the scaling parameter is used to characterize the correlation between the current attribute and the reference attribute. The inverse transformation module is used to perform an inverse transformation based on the first attribute prediction value a