Adaptive Loop Filtering with Non-Local Taps for Video Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems, such as VVC, face challenges in effectively reducing video reconstruction impairments due to the limitations of conventional Adaptive Loop Filters (ALF) in handling complex video sources, particularly in non-local and high-degree tap applications.
Innovation Solution
The introduction of an Adaptive Loop Filter (ALF) with non-local and high-degree taps that utilize additional filter inputs from neighboring blocks and reconstructed residual samples, enhancing the filtering process by incorporating pre-SAO samples and fixed filter outputs, and applying geometric transformations to improve video quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional Adaptive Loop Filter is used in video coding, then the system complexity is low and ease of manufacture is good, but video quality and coding efficiency are insufficient for complex video sources
Solution Approach 1:
The patent segments the filtering process by introducing multiple tap types (local taps, non-local taps, high-degree taps) that can be selectively applied to different regions and scenarios. This allows the system to divide the video content into different classification categories and apply appropriate filter strengths and tap configurations to each, thereby improving video quality without uniformly increasing complexity across all blocks
Solution Approach 2:
The patent implements dynamic adaptation by using block classification to determine which taps to apply and with what strength. The filter dynamically adjusts its behavior based on the classified characteristics of each block, selecting from multiple fixed filter sets and configuring tap applications adaptively. This dynamic approach allows the system to achieve high video quality for complex content while maintaining lower complexity for simpler content
2Productivity
If non-local and high-degree taps are added to ALF, then video quality and coding efficiency are improved, but computational complexity and processing time increase
Solution Approach 1:
The patent performs block classification and determines the appropriate filter configuration (which taps to apply, filter strength, etc.) before actually applying the filtering operation. This preliminary classification step allows the system to pre-determine the computational requirements for each block, enabling more efficient processing by avoiding unnecessary computations for blocks that don't require complex filtering
Solution Approach 2:
The patent changes the parameters of the filtering operation dynamically based on block characteristics. By adjusting which taps are applied (local vs. non-local, degree of taps), the filter strength, and selecting from multiple fixed filter sets, the system optimizes the balance between coding efficiency and processing time for different types of video content
3Reliability
If conventional ALF with limited taps is used, then processing speed is fast, but reconstruction impairments are not effectively reduced
Solution Approach 1:
The patent introduces non-local taps that reference samples from other blocks in the picture, acting as intermediaries to provide additional contextual information for the filtering process. These non-local taps serve as mediators that bring in relevant information from distant regions, improving reconstruction quality by considering broader spatial context without requiring direct adjacent samples
Data Source
AI summary
Method and apparatus for video coding using non-local, high-degree or residual taps. According to one method, a current filtered output is derived from an ALF for a current sample in the current block. The ALF comprises at least one non-local input term corresponding to a pre-SAO (Sample Adaptive Offset) reconstructed sample of the current sample and/or one or more neighbouring samples of the current sample, or a filter output, from one of a fixed filter sets, of the current sample and/or said one or more neighbouring samples of the current sample. According to another method, the ALF comprises at least one reconstructed sample from a non-adjacent block. According to yet another method, the ALF comprises at least one residual input term.