3D Video Decoding with Neural Post-Filter Interoperability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for applying neural network post filters on V3C occupancy, geometry, and attribute video streams lack sufficient procedures or restrictions to ensure interoperability between encoder and decoder devices and bitstreams.
Innovation Solution
Implementing a neural network post filter process specifically designed for V3C occupancy, geometry, and attribute video streams to enhance interoperability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If neural network post filter is applied on V3C occupancy, geometry, and attribute video streams, then quality of the video streams is improved, but interoperability between encoder and decoder devices is not guaranteed
Solution Approach 1:
The patent applies parameter changes by defining specific constraints and procedures for neural network post-filter application on V3C video streams. It establishes precise control parameters including filtering strength values (rsf_occupancy, rsf_geometry, rsf_attribute), filter activation flags, and processing conditions that must be consistently implemented by both encoders and decoders to ensure interoperability while maintaining quality improvements
Solution Approach 2:
The patent segments the neural network post-filter application into distinct processing stages for different video stream types (occupancy, geometry, attribute). Each stream type receives targeted filtering with specific parameters and conditions, allowing independent optimization and control of each component while maintaining overall system interoperability through standardized processing frameworks
2Reliability
If neural network post filter process is implemented on V3C video streams, then conformance is achieved, but device complexity increases
Solution Approach 1:
The patent uses copying by defining standardized parameter sets and filter configurations that can be replicated across different encoder and decoder implementations. The specified filtering strength values, activation flags, and processing procedures serve as templates that ensure consistent behavior across diverse devices, achieving conformance without requiring complex custom implementations
Solution Approach 2:
The patent creates a universal neural network post-filter framework that handles multiple video stream types (occupancy, geometry, attribute) through a single standardized processing mechanism. This multi-functional approach allows the same decoder architecture to process different stream types with appropriate parameter adjustments, reducing overall device complexity while maintaining broad conformance
Data Source
AI summary
A device may comprise one or more processors configured to: derive variables BitDepthY and BitDepthC equal to BitDepthY and BitDepthC; decode a slice QP SliceQPY and a offset QpBdOffsetY; derive StrengthControlVal equal to a value of (SliceQpY+QpBdOffsetY)÷(51+QpBdOffsetY) of a first slice of current coded picture; and set StrengthControlVal to an input tensor.