3D Data Refinement Processing for Encoding Distortion Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D data encoding and decoding methods using video encoding schemes cannot perform refinement processing on the relationship between occupancies, geometries, and attributes, limiting the ability to improve image quality through filter processing across multiple images/videos.
Innovation Solution
Incorporating a refinement information decoder and encoder to decode and encode refinement characteristics and activation information, allowing for refinement processing on attribute or geometry frames based on decoded characteristics, thereby enhancing image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If filter processing is performed only on attributes using neural network post-filter, then attribute refinement is achieved, but refinement using information between images (geometry and occupancy) is not possible
Solution Approach 1:
The neural network post-filter is extended to perform multiple functions: it now processes not only attribute data but also geometry and occupancy data. The system uses a single refinement processing unit that can apply neural network filtering across different data types (attribute, geometry, occupancy), making the filter universal and enabling comprehensive refinement using inter-image information.
Solution Approach 2:
The refinement processing is divided into separate neural network models for different data types: one for attribute refinement, another for geometry refinement, and a third for occupancy refinement. Each neural network is specifically designed to process its corresponding data type while utilizing information from other data types, allowing specialized refinement for each component while maintaining overall system integration.
2Manufacturing precision
If refinement processing is added to handle relationships between occupancies, geometries, and attributes, then image quality improvement is enabled, but encoding complexity increases
Solution Approach 1:
The refinement information encoder and decoder are merged into the existing encoding and decoding apparatus structures. The refinement characteristics information and activation information are integrated into the bitstream along with other encoding data, and the refinement information decoder is combined with the attribute decoder, geometry decoder, and occupancy decoder in a unified decoding architecture, reducing overall system complexity despite adding refinement functionality.
Solution Approach 2:
Refinement characteristics information and activation information serve as intermediary data structures that bridge the existing encoders/decoders and the new refinement processing unit. These intermediary elements carry the necessary control signals and parameters without requiring fundamental changes to the core encoding/decoding architecture, thus managing complexity while enabling enhanced functionality.
Data Source
AI summary
A 3D data decoding apparatus includes a geometry decoder configured to decode a geometry frame from encoded data, an attribute decoder configured to decode an attribute frame from the encoded data, a refinement information decoder configured to decode refinement characteristics information and refinement activation information from the encoded data, and a refinement processing unit configured to perform refinement processing on the attribute frame or the geometry frame according to the refinement characteristics information. The 3D data decoding apparatus further includes a syntax element indicating a number of refinements is decoded from the refinement activation information and an index indicating the refinement characteristics information for the decoded number of refinements is decoded.