Fractional-Pel Block Vectors for More Accurate Video Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques, such as HEVC and VVC, limit coding gains for non-screen content by using integer-pel resolution block vectors, which restrict the application of Intra Block Copy (IBC) and Intra Template Matching (IntraTMP) modes.
Innovation Solution
Implementing block vectors with fractional-pel resolution that identify positions between samples in the current picture, allowing for improved prediction accuracy and coding efficiency for both natural and screen content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If integer-pel resolution block vectors are used, then device complexity is reduced and ease of operation is improved, but coding efficiency and prediction accuracy deteriorate for non-screen content
Solution Approach 1:
The patent implements dynamic block vector resolution by allowing the block vector precision to be adaptively selected between integer-pel and fractional-pel resolutions based on the content type and coding conditions. This enables the system to switch between coarse (integer-pel) and fine (fractional-pel) precision modes, optimizing both computational efficiency and prediction accuracy for different video content characteristics
Solution Approach 2:
The patent changes the resolution parameter of block vectors from fixed integer-pel to variable precision including fractional-pel (e.g., 1/4-pel, 1/8-pel). This parameter change allows the block vector to identify positions between sample points in the reference picture, thereby improving prediction accuracy for natural content while maintaining compatibility with existing video coding frameworks
2Manufacturing precision
If fractional-pel resolution block vectors are implemented, then coding efficiency and prediction accuracy are improved, but device complexity increases
Solution Approach 1:
The patent segments the block vector precision requirement by applying different resolutions (integer-pel or fractional-pel) to different blocks or regions based on their content characteristics. This segmentation allows fractional-pel precision to be applied only where necessary for natural content, while integer-pel suffices for screen content, thereby reducing overall computational complexity
Solution Approach 2:
The patent creates a universal block vector mechanism that can function at multiple precision levels (integer-pel and fractional-pel) within the same video coding system. This multi-functionality allows the system to handle both screen content and natural content efficiently without requiring separate coding paths, thus managing complexity while maintaining versatility
3Productivity
If fractional-pel resolution block vectors are used, then coding efficiency is improved for natural content, but loss of information increases due to interpolation requirements
Solution Approach 1:
The patent introduces interpolation filters as intermediary mechanisms that generate fractional-pel sample values between integer sample points in the reference picture. These interpolated values serve as mediators that enable precise block vector positioning without directly accessing non-existent sample data, thereby improving prediction accuracy while managing the information loss through controlled interpolation
Data Source
AI summary
A method of encoding or decoding video data includes determining a block vector for a current block of a current picture, the block vector identifying a fractional-pel position between reference samples in the current picture and having a fractional-pel resolution that is less than integer-pel resolution; generating a prediction block from reference samples in the current picture based on the fractional-pel position; and encoding or decoding the current block based on the prediction block.