Contour Prediction Mode for Video Block Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern video codecs face inefficiencies in coding larger blocks with non-square shapes, particularly in angular prediction modes, leading to higher prediction errors and increased signaling costs, which negatively impact coding efficiency.
Innovation Solution
The introduction of a contour prediction mode that varies the prediction direction across a block by using offset values for each line or group of lines, allowing for more precise prediction of larger blocks with less prediction error and reduced signaling costs, achieved by determining offset values from a set of predetermined values and a global offset value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If angular prediction mode is used for larger blocks, then prediction coverage is improved, but prediction accuracy deteriorates due to non-linear contours
Solution Approach 1:
The prediction block is divided into multiple line units, with each line having its own offset value that can be independently determined. This segmentation allows the prediction direction to vary across different lines, capturing non-linear contours while maintaining large block coverage.
Solution Approach 2:
The prediction mode transitions from a static single offset value for the entire block to a dynamic configuration where each line can have a different offset value. This allows the prediction direction to adapt to local contour variations, improving accuracy for non-linear structures.
2Manufacturing precision
If contour prediction mode with multiple offset values is used, then prediction accuracy is improved, but signaling cost increases
Solution Approach 1:
Instead of applying a uniform high-precision prediction mode across the entire block, the solution applies different offset values locally to each line only when needed. This allows high prediction accuracy for lines with non-linear contours while using simpler prediction for other lines, reducing overall signaling cost.
Solution Approach 2:
The patent changes the parameter representation from a single block-level offset value to multiple line-level offset values. This parameter transformation enables more precise prediction by capturing local variations, while the ability to selectively apply different parameters to different lines helps control signaling overhead.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A video decoder (20) for supporting a prediction mode for predicting blocks (2) of a video, wherein the video decoder (20) is configured to predict each of the blocks (2) by extrapolating a neighborhood of the respective block (2) into the block along a direction which varies across the respective block.