Adaptive Loop Filter Padding for Video Coding Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in efficiently handling virtual boundaries of coding tree blocks, leading to suboptimal compression efficiency and increased complexity in coding and decoding processes.
Innovation Solution
The proposed solution involves methods for video encoders and decoders to manage virtual boundaries by determining the availability of samples outside current blocks and using padded samples or virtual samples for adaptive loop filter processes, thereby simplifying implementations and improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If video coding standards use traditional boundary handling methods for coding tree blocks, then the implementation is simpler, but compression efficiency is suboptimal
Solution Approach 1:
The patent segments the video picture into coding tree blocks (CTBs) and further divides them into coding tree units (CTUs) with virtual boundaries. By introducing virtual boundaries within CTBs, the filtering process can be independently applied to each unit while maintaining overall compression efficiency. This segmentation allows parallel processing and reduces the complexity of handling entire picture boundaries.
Solution Approach 2:
The patent introduces virtual samples as intermediary elements that facilitate the adaptive loop filter process across block boundaries. These virtual samples act as mediators that enable filtering operations to extend beyond actual picture boundaries without requiring complex boundary checking logic, thus improving compression efficiency while maintaining implementation simplicity.
2Ease of operation
If samples outside current blocks are made unavailable for coding process, then boundary handling is simpler, but adaptive loop filter performance deteriorates
Solution Approach 1:
The patent performs preliminary action by generating virtual samples before the adaptive loop filter process. These virtual samples are pre-computed based on available neighboring samples and stored in a buffer, allowing the filtering process to proceed without complex boundary checks during actual filtering operations. This preliminary preparation maintains simplicity while ensuring filter performance.
Solution Approach 2:
The patent creates copies of existing samples to generate virtual samples. By copying and transforming available neighboring samples (using techniques like reflection, repetition, or gradient extension), the system creates virtual samples that maintain the statistical properties of real samples, thereby preserving adaptive loop filter performance without accessing actual boundary samples.
3Productivity
If virtual boundaries of coding tree blocks are not properly handled, then coding process is simpler, but compression efficiency and filtering accuracy are reduced
Solution Approach 1:
The patent introduces a virtual boundary dimension within the coding tree block structure. By defining virtual boundaries at sub-CTB levels (e.g., every 4 or 8 samples), the system creates an additional hierarchical layer that enables finer-grained filtering operations. This dimensional addition allows independent filtering units without requiring complex management of traditional picture-level boundaries.
Solution Approach 2:
The patent changes the parameter of boundary location from fixed picture boundaries to flexible virtual boundaries within CTBs. By making the boundary position a configurable parameter (e.g., offset values indicating virtual boundary locations), the system can adaptively adjust filtering unit sizes and positions, improving compression efficiency while managing complexity through parameterized control rather than hard-coded boundary logic.
Data Source
AI summary
An example method of video processing includes determining, for a conversion between a current block of a video and a bitstream representation of the video, that one or more samples of the video outside the current block are unavailable for a coding process of the conversion. The coding process comprises an adaptive loop filter (ALF) coding process. The method also includes performing, based on the determining, the conversion by using padded samples for the one or more samples of the video. The padded samples are generated by checking for availability of samples in an order.