Transform Domain Prediction for Non-Square Video Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression technologies face challenges in efficiently processing next-generation video content with high spatial resolution, high frame rate, and high dimensionality, particularly in accurately predicting and encoding video signals in the frequency domain.
Innovation Solution
The method involves applying correlations between frequency coefficients when a prediction block is split, using correlation coefficients or scaling coefficients to regenerate prediction blocks in the frequency domain, and designing prediction filters to improve coding efficiency and prediction performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a prediction block is split into multiple prediction units, then the prediction accuracy for different regions can be improved, but the complexity of the prediction process increases
Solution Approach 1:
The prediction block is divided into multiple prediction units (PUs) that can be processed independently. Each PU can have its own motion vectors and prediction parameters, allowing region-specific optimization while maintaining overall prediction accuracy. This segmentation enables better handling of non-uniform motion patterns without requiring complete redesign of the prediction architecture.
Solution Approach 2:
Different prediction units within the same prediction block are allowed to have different prediction parameters, motion vectors, and correlation coefficients. This local quality approach enables each PU to be optimized for its specific characteristics while the overall prediction block maintains coherence through shared context information.
2Productivity
If transform domain prediction is applied to the entire prediction block, then processing efficiency is improved, but prediction accuracy decreases when the block is split
Solution Approach 1:
The transform domain prediction is segmented to operate on individual prediction units rather than the entire prediction block. Each PU undergoes independent transform and correlation coefficient application, allowing efficient processing while maintaining accuracy through localized optimization. The segmented approach enables parallel processing of multiple PUs.
Solution Approach 2:
The prediction system dynamically adapts the transform domain processing based on the prediction unit structure. When prediction blocks are split into multiple PUs, the system applies transform and correlation operations at the PU level rather than uniformly at the block level, optimizing both efficiency and accuracy through adaptive processing granularity.
3Measurement precision
If correlation coefficients are applied to all frequency components, then prediction quality is improved, but computational complexity increases
Solution Approach 1:
Different correlation coefficients are applied to different frequency components based on their specific characteristics. The system calculates and applies correlation coefficients selectively for each frequency component or group of components, rather than using a uniform correlation approach for all frequencies. This local quality approach optimizes prediction quality while managing computational complexity through targeted processing.
Data Source
AI summary
The present invention provides a method for decoding a video signal including extracting an inter prediction mode of a current block from the video signal; generating two non-square prediction blocks using two motion vectors according to the inter prediction mode; performing a non-square transform on the two non-square prediction blocks to obtain two non-square prediction blocks on a frequency domain; updating the two non-square prediction blocks on the frequency domain using a correlation coefficient or a scaling coefficient; and generating a reconstruction block based on the updated non-square prediction blocks on the frequency domain and a residual block.