Adaptive Loop Filter Allocation for Lower-Complexity Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies like VVC require iterative processing to determine a suitable filter set for each block, which increases implementation costs and affects compression efficiency.
Innovation Solution
A video coding apparatus that acquires statistical information on a picture and allocates a single filter set common to multiple blocks, using this information to perform adaptive loop filter processing without iterative calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a suitable filter set is determined for each block using iterative processing, then prediction accuracy is improved, but implementation cost increases
Solution Approach 1:
The patent merges the filter selection process from block-level to picture-level by allocating a single filter set common to all blocks. This is achieved by acquiring statistical information at the picture level and using it to select one filter set that will be applied uniformly across all blocks, thereby eliminating the need for iterative processing per block while maintaining overall prediction accuracy.
Solution Approach 2:
The patent makes the filter set universal across all blocks in the picture by selecting a single filter set that serves multiple blocks. The allocation unit allocates this universally applicable filter set based on picture-level statistical information, allowing the same filter set to be used for filtering multiple different blocks without requiring block-specific optimization.
2Productivity
If multiple candidate filter sets are provided for adaptive loop filter, then compression efficiency is improved, but processing complexity increases
Solution Approach 1:
The patent extracts the filter selection decision from the block-level processing stage and moves it to the picture-level preparation stage. By acquiring statistical information and selecting the filter set at the picture level before block processing, the patent removes the iterative selection complexity from each block while retaining the benefit of having multiple candidate filter sets available at the picture level.
Solution Approach 2:
The patent performs filter set selection in advance at the picture level before actual block filtering occurs. The allocation unit pre-selects a suitable filter set based on picture-level statistical information, so that when block filtering is performed, the filter set is already determined and ready for use, eliminating the need for iterative processing during block-level operations.
3Measurement precision
If iterative processing is performed to evaluate difference cost, then filter selection accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs the filter selection action preliminarily at the picture level before block processing. By acquiring statistical information and selecting the filter set in advance based on picture-level characteristics, the patent eliminates the need for time-consuming iterative difference cost evaluation during block processing, significantly reducing processing time while maintaining filter selection accuracy.
Solution Approach 2:
The patent performs filter selection at the picture level which is a coarser granularity than block-level selection. This partial action approach selects a filter set that is suitable for the overall picture characteristics, sacrificing some block-specific optimization but gaining significant processing time efficiency by avoiding iterative evaluation for each block.
Data Source
AI summary
In order to achieve improved compression efficiency while controlling implementation costs, an adaptive loop filter includes an acquisition unit configured to acquire statistical information on a picture to be coded, an allocation unit configured to allocate a single filter set common to a plurality of blocks included in the picture, based on the statistical information on the picture, and a filter processing unit configured to execute adaptive loop filter processing on each of the plurality of blocks, by using the single filter set.