Mesh Residual Context Modeling Using Predictor Indexes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video-based dynamic mesh coding (V-DMC) technologies face inefficiencies in residual bit estimation and encoding due to the limited use of residual context models, particularly in distinguishing between geometric and attribute residuals.
Innovation Solution
Enhance the residual context model by incorporating a predictor index (predIndex) as an additional dimension, allowing for improved separation of residual classes and probability updates in both geometry and attribute contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the residual context model uses limited dimensions (2D for geometry, 1D for attribute), then the model complexity is kept low, but the encoding efficiency and compression performance deteriorate
Solution Approach 1:
The patent applies dimensionality change by adding the predictor index (predIndex) as a new dimension to the residual context model. For geometry residuals, it transitions from a 2D model (ctxIndex, residueGroup) to a 3D model (ctxIndex, residueGroup, predIndex). For attribute residuals, it transitions from a 1D model (ctxIndex) to a 2D model (ctxIndex, predIndex). This additional dimension enables more fine-grained probability estimation and improves encoding efficiency by better distinguishing between different residual types.
2Measurement precision
If the residual context model distinguishes between geometry and attribute residuals using separate models, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The patent applies segmentation by creating separate residual context models for geometry residuals and attribute residuals. The geometry residual model uses dimensions (ctxIndex, residueGroup, predIndex) while the attribute residual model uses dimensions (ctxIndex, predIndex). This segmentation allows each model to be optimized for its specific residual type, improving measurement precision in residual classification and probability estimation.
Data Source
AI summary
A method performed by at least one processor in an encoder includes receiving a polygon mesh comprising a plurality of vertices; predicting a vertex using an original vertex from the plurality of vertices; determining a residual between the predicted vertex and the original vertex; determining a residual context from a plurality of residual contexts; encoding the residual between the predicted vertex and the original vertex using the residual context; generating a bitstream including (i) the encoded residual, and (ii) a plurality of indexes identifying the residual context including (a) three indexes from the plurality indexes when the residual is associated with a geometric position of the predicted vertex, and (b) two indexes from the plurality of indexes when the residual is associated with an attribute of the predicted vertex.