Adaptive In-Loop Filter with Multiple Classifications
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Wiener-based adaptive filters in video coding use a single classifier for in-loop filtering, which may not be optimal for all images, limiting the reduction of errors between original and reconstructed pictures and thus affecting compression performance.
Innovation Solution
Implementing a system that selects the best classification from multiple available classifications based on error correction effectiveness, using a combination of feature descriptors like Laplacian, sample-based, and ranking-based descriptors to classify samples and apply corresponding filters, thereby improving the compression performance marginally increasing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a single classifier is used for in-loop filtering, then device complexity is reduced, but manufacturing precision (filtering accuracy) deteriorates
Solution Approach 1:
The patent divides the filtering process into multiple segments by using different classifiers for different image regions. Instead of applying a single classifier uniformly, the system segments the image into regions with different characteristics and applies appropriate classifiers to each segment, thereby improving overall filtering accuracy while managing complexity through selective application.
Solution Approach 2:
The patent implements dynamic classifier selection where the system adaptively chooses which classifier to apply based on the local characteristics of each image region. This dynamic approach allows the filtering process to respond to varying image content, improving accuracy by matching classifiers to appropriate regions while maintaining reasonable complexity through conditional application rather than exhaustive processing.
2Manufacturing precision
If multiple feature-based classifications are implemented, then filtering accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies local quality by using different feature descriptors and classification approaches for different regions of the image based on their specific characteristics. Rather than uniformly applying complex multiple feature-based classification to the entire image, the system identifies regions where such classification provides the most benefit and applies it selectively, thereby improving filtering accuracy in critical areas while minimizing overall processing time.
Data Source
Figure 1
Figure 2
Figure 2A
AI summary
A decoder is described, which receives an encoded data stream. The encoded data stream includes data representing a picture and a classification mode parameter. The classification mode parameter specifies a certain classification to be applied to samples in the picture, the certain classification selected from a plurality of classifications. The decoder decodes the encoded data stream to obtain a reconstructed picture and the classification mode parameter, classifies samples of the reconstructed picture into one of a plurality of classes using the certain classification specified by the classification mode parameter, and filters the samples of the reconstructed picture. A sample is filtered using a filter for the class into which the sample has been classified.