V3C Patch Remeshing for Volumetric Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current volumetric video coding standards, such as V3C, face challenges in efficiently compressing and reconstructing three-dimensional mesh data, leading to increased processing time and memory usage due to the generation of excessive faces and vertices during the decoding and rendering of volumetric video, which affects the quality and efficiency of AR, VR, and MR applications.
Innovation Solution
The implementation of a V3C patch remeshing method that subsamples geometry components, selects salient points, triangulates these points, and iteratively refines the mesh to minimize the number of triangles and vertices while maintaining quality, allowing for adaptive level-of-detail and parallel processing, thereby reducing computational complexity and improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional volumetric video coding methods are used to encode three-dimensional mesh data, then the complete geometric information is preserved, but the number of faces and vertices increases excessively, leading to increased processing time and memory usage
Solution Approach 1:
The patent extracts only the essential geometric features by identifying and retaining salient points (extrema, umbilical points, texture singularities) while removing redundant vertices and faces. This extraction process maintains the critical shape characteristics needed for accurate reconstruction while significantly reducing the total number of mesh elements, thereby decreasing processing time without completely sacrificing geometric fidelity.
Solution Approach 2:
The patent applies different levels of mesh density to different regions of the three-dimensional object based on local geometric importance. Salient points and regions with high curvature or texture significance retain higher mesh density, while flat or less important regions use coarser sampling. This local quality approach preserves essential geometric information in critical areas while reducing overall vertex and face counts to improve processing efficiency.
2Measurement precision
If traditional volumetric video coding methods are used to encode three-dimensional mesh data, then the complete geometric information is preserved, but the memory usage increases due to excessive faces and vertices
Solution Approach 1:
The patent extracts only the essential geometric features by identifying and retaining salient points (extrema, umbilical points, texture singularities) while removing redundant vertices and faces. This extraction process maintains the critical shape characteristics needed for accurate reconstruction while significantly reducing the total number of mesh elements, thereby decreasing processing time without completely sacrificing geometric fidelity.
Solution Approach 2:
The patent applies different levels of mesh density to different regions of the three-dimensional object based on local geometric importance. Salient points and regions with high curvature or texture significance retain higher mesh density, while flat or less important regions use coarser sampling. This local quality approach preserves essential geometric information in critical areas while reducing overall vertex and face counts to improve processing efficiency.
3Productivity
If the number of triangles and vertices is reduced through subsampling, then processing time and memory usage decrease, but the quality of the reconstructed three-dimensional object may deteriorate
Solution Approach 1:
The patent applies different levels of mesh density to different regions of the three-dimensional object based on local geometric importance. Salient points and regions with high curvature or texture significance retain higher mesh density, while flat or less important regions use coarser sampling. This local quality approach preserves essential geometric information in critical areas while reducing overall vertex and face counts to improve processing efficiency.
Solution Approach 2:
The patent incorporates an evaluation mechanism that assesses the quality of depth patch triangulation by comparing the reconstructed three-dimensional object against the original. This feedback loop allows the system to iteratively adjust subsampling parameters and salient point selection to achieve the optimal balance between processing efficiency and reconstruction quality, ensuring that quality degradation is minimized while maintaining reduced processing requirements.
4Ease of manufacture
If regular triangulation is applied to all geometry components, then the processing method is simple, but the computational complexity increases due to excessive triangles
Solution Approach 1:
The patent applies different levels of mesh density to different regions of the three-dimensional object based on local geometric importance. Salient points and regions with high curvature or texture significance retain higher mesh density, while flat or less important regions use coarser sampling. This local quality approach preserves essential geometric information in critical areas while reducing overall vertex and face counts to improve processing efficiency.
Solution Approach 2:
The patent changes key parameters of the triangulation process by introducing subsampling rates and salient point selection criteria. Instead of uniformly triangulating all geometry components, the system adjusts the density and distribution of triangles based on local geometric features, thereby reducing the total number of triangles and computational complexity while maintaining processing simplicity through standardized algorithms adapted to local requirements.
Data Source
AI summary
An apparatus comprising circuitry configured to: obtain a three-dimensional model comprising at least one patch, at least one geometry component, at least one occupancy component, and zero or more texture components; subsample the at least one geometry component of the at least one patch of a three-dimensional object at occupied positions using a subsampling criterion; define respective search windows around the respective occupied positions; select respective salient points relative to the respective occupied positions within the respective search windows; triangulate the salient points to approximate a shape of the three-dimensional object; detect zero or more triangles that overlap with at least one unoccupied pixel; split the zero or more triangles that overlap with at least one unoccupied pixel until no triangle overlaps with the unoccupied pixels; and add zero or more additional triangles close to a border of the three-dimensional object to generate a resulting mesh signaled to a decoder.