Adaptive Orthogonal Transform Selection for Chroma Prediction Residuals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing cross-component intra prediction modes, such as CCIP_A and CCIP_L, suffer from biased energy distribution trends in chroma blocks, leading to reduced encoding efficiency due to uneven prediction accuracy across different areas of the block, as they rely on neighboring decoded pixels for calculating linear models.
Innovation Solution
An image encoding and decoding device that includes a cross-component intra predictor and a candidate generator to select an optimal orthogonal transform type based on the positions of neighboring decoded pixels, allowing for adaptive orthogonal transform processing to match the energy distribution trends of prediction residuals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cross-component intra prediction modes (CCIP_A, CCIP_L) are used to predict chroma blocks by referring to neighboring decoded pixels, then prediction accuracy in specific areas is improved, but encoding efficiency deteriorates due to biased energy distribution trends
Solution Approach 1:
The patent applies local quality by making the orthogonal transform type adaptive to different regions and prediction modes. Specifically, when CCIP_A mode is used (referring only to top neighboring pixels), DST-7 transform is selected to handle the vertical bias in energy distribution. When CCIP_L mode is used (referring only to left neighboring pixels), DCT-8 transform is selected to handle the horizontal bias. This local adaptation of transform types to match local prediction characteristics resolves the contradiction between prediction accuracy and encoding efficiency.
Solution Approach 2:
The patent changes the parameter of orthogonal transform type based on the prediction mode and neighboring pixel configuration. By selecting different transform types (DST-7 for vertical bias, DCT-8 for horizontal bias) according to the specific prediction scenario, the system optimizes energy concentration and encoding efficiency while maintaining prediction accuracy.
2Device complexity
If a fixed orthogonal transform type is used for all prediction modes, then device complexity is reduced, but encoding efficiency deteriorates due to inability to match energy distribution trends
Solution Approach 1:
The patent introduces dynamics by making the orthogonal transform type selection adaptive rather than fixed. The transform type is dynamically chosen based on the prediction mode (CCIP_A or CCIP_L) and the configuration of neighboring decoded pixels. This dynamic adaptation allows the system to match energy distribution trends in different scenarios, improving encoding efficiency without significantly increasing device complexity through simple conditional selection logic.
3Measurement precision
If cross-component intra prediction uses both top and left neighboring decoded pixels, then prediction accuracy is improved, but a correct linear model cannot be calculated when luminance-chroma relationships differ between top and left sides
Solution Approach 1:
The patent applies segmentation by dividing the prediction approach into separate modes: CCIP_A uses only top neighboring pixels, while CCIP_L uses only left neighboring pixels. This segmentation allows the system to select the appropriate prediction mode based on which neighboring side has a more reliable luminance-chroma relationship, avoiding the problem of conflicting relationships when both top and left sides are used together.
Data Source
AI summary
An image encoding device (1) includes the cross-component intra predictor (171a) generates a predicted chroma block through cross-component intra prediction in which a chroma block to be encoded is predicted by referring to, as the neighbouring decoded pixels adjacent to the chroma block, decoded luminance pixels and decoded chroma pixels, a candidate generator (181) configured to generate candidates for an orthogonal transform type to be applied to orthogonal transform processing on prediction residuals that represent errors between the predicted chroma block and the chroma block; and a transformer (121) configured to perform the orthogonal transform processing on the chroma prediction residuals by using an orthogonal transform type selected from among the candidates generated by the candidate generator (181). The candidate generator (181) generates the candidates for the orthogonal transform type, depending on whether or not positions of the neighbouring decoded pixels referred to in cross-component intra prediction are to only any one of a top, a bottom, a left, and a right of a chroma block.