Adaptive Multiple Transform Selection for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques, such as H.264/AVC, HEVC, and VVC, face inefficiencies as image size, resolution, and frame rate increase, leading to higher data amounts that require more hardware resources and do not effectively utilize multiple transforms for residual signals, necessitating a method for improved coding efficiency and image enhancement.
Innovation Solution
A video coding method and apparatus that adaptively select multiple transform kernels based on the features of residual signals, forming multiple-transform-selection groups and selecting the optimal transform kernel for efficient transformation of residual blocks, thereby enhancing video quality and coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single transform kernel is used for residual signals, then device complexity is reduced, but coding efficiency deteriorates
Solution Approach 1:
The patent implements dynamic transform kernel selection by classifying residual signal types (e.g., vertical edges, horizontal edges, flat regions) and applying different transform kernels (DST, DCT, identity) based on the classified signal characteristics. This dynamic adaptation allows the system to optimize coding efficiency for each block while maintaining manageable device complexity through structured classification logic.
Solution Approach 2:
The patent changes the transform kernel parameter based on residual signal features. By analyzing signal characteristics and selecting from multiple transform kernels (DST-II, DCT-II, identity transform), the system adapts the transformation parameters to match the actual signal content, thereby improving compression efficiency without requiring a single complex transform for all cases.
2Productivity
If multiple transform kernels are used for residual signals, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments the residual signal processing by dividing the transform selection into distinct categories based on signal characteristics (vertical edges, horizontal edges, flat regions). Each segment uses a predetermined transform kernel, avoiding the need to evaluate all possible transforms for every block. This segmentation reduces device complexity while maintaining the benefits of multiple transforms.
Solution Approach 2:
The patent applies different transform kernels to different regions or types of residual signals based on their local characteristics. For example, DST kernels are applied to vertical edge regions, DCT kernels to horizontal edge regions, and identity transforms to flat regions. This localized adaptation improves coding efficiency for each region while keeping the overall system complexity manageable through targeted transform selection.
3Productivity
If transform selection is performed without signal feature analysis, then processing speed is improved, but coding efficiency deteriorates
Solution Approach 1:
The patent performs preliminary classification of residual signal types before applying transforms. By analyzing signal features (edge directions, flatness) in advance and selecting the appropriate transform kernel beforehand, the system avoids performing complex transform evaluations during the actual transformation process. This preliminary action maintains processing speed while enabling efficient transform selection for improved coding efficiency.
Data Source
AI summary
A method and an apparatus are disclosed for video using adaptive multiple transform selection. In the disclosed embodiments, a video decoding device obtains a multiple-transform-selection index and inverse-quantized residual signals of the current block from a bitstream. The video decoding device composes a plurality of multiple-transform-selection groups for the current block, and selects a multiple-transform-selection group from the multiple-transform-selection groups. The video decoding device derives a transform kernel from the multiple-transform-selection group by using the multiple-transform-selection index. The video decoding device generates inverse-transformed residual signals by inversely transforming the inverse-quantized residual signals by using the derived transform kernel.