Video Coding OOB Checks for Accurate Bi-Directional Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face inefficiencies due to out-of-bound (OOB) reference blocks during bi-directional motion compensation, leading to reduced prediction accuracy and increased computational complexity.
Innovation Solution
Implementing out-of-bound (OOB) checks to determine if reference blocks are outside the boundaries of reference pictures, and adjusting the prediction process accordingly by excluding OOB samples, using representative pixel positions for sub-blocks, and deriving motion vectors from multiple vectors to handle OOB conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bi-directional motion compensation is performed without OOB checks, then computational complexity is reduced, but prediction accuracy deteriorates due to use of invalid OOB samples
Solution Approach 1:
The patent performs OOB checks before the bi-directional motion compensation process to identify and exclude invalid reference samples in advance. This preliminary action prevents the use of OOB samples in prediction, improving accuracy without requiring complex runtime checks during the compensation process itself.
Solution Approach 2:
The patent extracts and removes OOB samples from the reference blocks before they are used in prediction. By identifying samples that fall outside picture boundaries and excluding them from the prediction process, the method ensures only valid samples contribute to the prediction, resolving the contradiction between using all samples (low complexity) and using only valid samples (high accuracy).
2Measurement precision
If OOB checks are performed on all reference blocks, then prediction accuracy is improved, but computational overhead increases
Solution Approach 1:
The patent applies OOB checks selectively to specific regions and samples within reference blocks rather than uniformly to all blocks. By performing checks only where boundary conditions are relevant and using representative pixel positions for sub-blocks, the method improves prediction accuracy in critical areas while minimizing unnecessary computational overhead in other areas.
Solution Approach 2:
The patent performs OOB checks at representative pixel positions for sub-blocks rather than every single pixel position. This partial action approach provides sufficient accuracy for prediction while significantly reducing the computational overhead compared to checking every sample, resolving the contradiction between thorough checking and computational efficiency.
Data Source
AI summary
A method of coding pixel blocks using out-of-bound (OOB) checks is provided. A video coder receives data to be encoded or decoded as a current block of a current picture of a video. The video coder identifies a first reference block in a first reference picture based on a first block vector of the current block. The video coder performs OOB check for the first reference block relative to a boundary of a sub-unit of the first reference picture. The video coder generates a predictor for the current block based on the first reference block and based on the OOB check. When a sample of the first reference block is OOB, the sample is not used in a bi-directional motion compensation for the current block. The video coder encodes or decodes the current block by using the generated predictor.


