Chrominance Prediction via Down-sampled Luminance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding schemes face inefficiencies in coding chrominance pixels from luminance pixels, particularly due to high data transfer loads in small blocks and large coding amounts for down-sampling filters, and in methods like FRUC and BTM, where motion vector estimation is computationally intensive.
Innovation Solution
A decoding and coding device that down-samples luminance images to generate prediction images for chrominance, using a predictor to derive down-sampled images and parameters based on position relationships between luminance and chrominance pixels, reducing data transfer and improving coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If luminance pixels are down-sampled to predict chrominance pixels, then coding efficiency is improved, but the coding amount of switching signals for selecting down-sampling filters increases
Solution Approach 1:
The patent changes the parameter representation by encoding the filter selection index using variable-length coding instead of fixed-length coding. This allows frequently used filters to be represented with fewer bits, reducing the overall coding amount of switching signals while maintaining the ability to select from multiple down-sampling filters for different chrominance prediction modes
2Measurement precision
If motion vector estimation is performed in FRUC and BTM schemes, then prediction accuracy is improved, but the data transfer load increases significantly in small blocks
Solution Approach 1:
The patent segments the motion estimation process by performing it only in the vertical direction for FRUC and in both directions for BTM, rather than performing full 2D motion estimation. This segmentation reduces the computational complexity and data transfer requirements while maintaining adequate prediction accuracy for the specific application scenarios
Solution Approach 2:
The patent applies different motion estimation strategies to different prediction modes: FRUC uses vertical-only motion estimation for inter-frame chrominance prediction, while BTM uses bidirectional motion estimation for intra-frame prediction. This localized approach optimizes the balance between prediction accuracy and data transfer load for each specific prediction scenario
Data Source
AI summary
A prediction image generation method is provided. The method derives a down-sampled luminance image and a down-sampled neighboring luminance image by down-sampling the luminance image of a target block and a neighboring luminance image based on information related to a down-sampling and indicating one of a plurality of position relationships between at least one luma pixel and a chroma pixel. The method derives a neighboring chrominance image, parameters derived from the down-sampled neighboring luminance image and the neighboring chrominance image, and the prediction image by using the down-sampled luminance image and the parameters. The position relationships include that a position of the chroma pixel is identical to a position of one of the at least one luma pixel and that the position of the chroma pixel is located in an intermediate of two of the at least one luma pixel on a left side of a 2×2 luma pixel block.