Neural Network Video Filter for Block Effect Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding and decoding systems suffer from block effects due to differences in coding parameters between neighboring Coding Units (CUs), leading to reduced subjective and objective quality of reconstructed pictures and impaired prediction accuracy.
Innovation Solution
A filtering method that acquires sample information and side information, and inputs multiple components of this information into a neural network-based filter to produce filtered components, thereby improving the quality of reconstructed pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a neural network-based filter with multiple inputs is used to process multiple color components, then the quality of reconstructed pictures is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the processing of multiple color components by using separate neural network filters for each component (Y, U, V channels) rather than processing them together. This segmentation reduces the computational complexity of each individual filter while still achieving improved picture quality through component-specific optimization.
Solution Approach 2:
The patent applies partial action by selectively processing only certain color components that need filtering, rather than uniformly processing all components. The filter is applied conditionally based on the specific picture block characteristics, reducing unnecessary computational operations while maintaining quality where needed.
2Productivity
If a pre-processing filter is used to decrease video resolution, then fewer bits are needed for representation and coding efficiency is improved, but the quality of the original picture is reduced
Solution Approach 1:
The patent applies preliminary action by performing filtering operations before the main encoding process. The pre-processing filter prepares the picture data in advance, reducing block effects and improving the quality of reference pictures, which subsequently improves coding efficiency without significantly compromising the perceived picture quality.
Solution Approach 2:
The patent converts the harmful block effects introduced by coding processes into a benefit by applying targeted filtering. The filter exploits the block effect patterns to improve reference picture quality, which indirectly benefits coding efficiency. The quality reduction is localized to areas where block effects are minimal, preserving overall picture quality while achieving coding gains.
3Manufacturing precision
If a post filter is used to process in-loop filtered video, then the resolution of the output video is improved, but the computational complexity increases
Solution Approach 1:
The patent applies local quality by differentiating the filtering strength and application based on the specific characteristics of each picture block. Areas with severe block effects receive stronger filtering, while areas with minimal block effects receive lighter or no filtering. This localized approach improves output quality where needed without unnecessarily increasing computational complexity in already-good regions.
Data Source
AI summary
Disclosed are a filtering method and device, an encoder and a computer storage medium. The method comprises: acquiring sample information to be filtered, acquiring at least one piece of side information, and inputting at least two components of the sample information to be filtered and at least one piece of side information into a filter based on a neural network so as to output at least one component after the sample information to be filtered is filtered. Further provided are a filtering device, an encoder and a computer storage medium.