Color-Sampling-Aware Pixel Prediction for Low-Memory Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video decoding methods require excessive internal memory usage due to the derivation of prediction samples, particularly in formats like YUV4:2:0, which complicates hardware implementation.
Innovation Solution
The method determines prediction samples by considering the color sampling mode of the picture block, ensuring consistency between the prediction sample and the block, thereby reducing the memory required for derivation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prediction samples are derived using conventional methods, then prediction accuracy is maintained, but internal memory usage increases excessively
Solution Approach 1:
The patent applies local quality by differentiating the derivation process based on color sampling modes. For YUV4:4:4 mode, full-resolution prediction samples are derived, while for YUV4:2:0 mode, a simplified derivation process is used that operates on downsampled chroma components. This localized adaptation of the derivation method to match the actual data resolution maintains prediction accuracy where needed while reducing memory usage in appropriate contexts.
Solution Approach 2:
The patent changes the parameter of color sampling mode (YUV4:4:4 vs. YUV4:2:0) to determine the derivation approach. By detecting the color sampling mode first, the system dynamically adjusts the prediction sample derivation process, using full-resolution references for 4:4:4 and downsampled references for 4:2:0, thereby optimizing the balance between accuracy and memory consumption based on the actual input characteristics.
2Loss of information
If conventional prediction sample derivation is used, then comprehensive pixel information is obtained, but hardware implementation complexity increases
Solution Approach 1:
The patent performs preliminary action by first determining the color sampling mode before initiating the prediction sample derivation process. This preliminary detection step allows the system to select the appropriate derivation path (full-resolution or downsampled) in advance, avoiding the need for complex hardware that must handle all possible cases simultaneously. The preliminary classification simplifies the subsequent derivation hardware by making the processing path deterministic based on the detected mode.
3Quantity of substance
If downsampled chroma components are used for prediction, then memory occupancy is reduced, but prediction precision may be affected
Solution Approach 1:
The patent applies dynamics by making the prediction sample derivation process adaptive rather than static. The system dynamically adjusts the derivation methodology based on the detected color sampling mode. For YUV4:2:0, it uses downsampled chroma components with an adjusted derivation process, while for YUV4:4:4, it uses full-resolution components. This dynamic adaptation ensures that the system uses the appropriate level of detail for each case, maintaining precision when available and reducing memory usage when appropriate.
Data Source
AI summary
A method for determining a prediction sample of a picture block includes acquiring a type of a current pixel string where a current pixel in a current picture block is located, and determining a reference pixel value of the current pixel according to the type of the current pixel string. The method further includes obtaining a prediction sample of the current pixel according to a color sampling mode of the current picture block and the reference pixel value of the current pixel. A color sampling mode of the obtained prediction sample is consistent with the color sampling mode of the current picture block. The method further includes obtaining prediction values of the current picture block according to the prediction sample of the current pixel.