3D Data Decoding Apparatus with Attribute Image Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D data coding schemes, such as those using HEVC or VVC, suffer from distortion and reduced image quality due to encoding of color images, and deep learning post-filters struggle to adjust filter strength effectively across varying image features in 3D data encoding and decoding.
Innovation Solution
A 3D data decoding apparatus and coding apparatus that incorporate a header decoder for supplemental enhancement information, an attribute image decoder, and an attribute image filter unit to derive and apply filter parameters, enabling precise filtering of attribute images based on objective information and neural network models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If color images are encoded using video coding schemes (HEVC/VVC), then compression efficiency is improved, but image quality and accuracy of reconstructed 3D data deteriorate due to encoding distortion
Solution Approach 1:
The patent separates the encoding process into two independent parts: geometry data is encoded using efficient video coding schemes (HEVC/VVC), while attribute images are encoded separately using a dedicated attribute image encoder. This segmentation allows each component to be optimized independently, maintaining compression efficiency for geometry while preserving attribute image quality.
Solution Approach 2:
The patent introduces an attribute image filter unit as an intermediary component that receives decoded attribute images and applies deep learning-based post-filtering. This intermediary process compensates for encoding distortion by learning the mapping between encoded and original attribute images, thereby recovering image quality without sacrificing compression efficiency.
2Manufacturing precision
If deep learning post-filter is applied with fixed filter strength, then image quality improvement is achieved, but adaptability to different image features and frames is reduced
Solution Approach 1:
The patent implements dynamic filter strength adjustment by deriving filter parameters from the actual attribute images themselves. The attribute image filter unit analyzes image features and automatically adjusts filtering strength on a per-frame and per-attribute-channel basis, allowing the system to adapt to varying image content while maintaining high quality improvement.
Solution Approach 2:
The patent changes the filter strength parameter from a fixed value to a dynamically derived parameter based on image characteristics. The attribute image filter unit modifies filtering parameters according to the specific features detected in each frame and attribute channel, enabling optimal filtering performance across diverse 3D data scenarios.
Data Source
AI summary
Use of only supplemental enhancement information of a deep learning post-filter cannot adjust a filter strength in a case that each frame has a different image feature, thus reducing image quality improvement performance. A decoding apparatus for decoding coded data includes a header decoder that decodes supplemental enhancement information of a deep learning post-filter from the coded data, an attribute image decoder that decodes an attribute image from the coded data, and an attribute image filter unit that performs filtering processing of the attribute image, wherein the header decoder decodes objective information indicating that the attribute image filter unit is to perform filtering processing on the attribute image.