Video Coding QP Adaptive Mixture-of-Experts Filter
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video processing technologies face challenges in optimizing picture quality, particularly in video codecs, due to limitations in intra prediction and motion compensation techniques, which affect compression efficiency and bandwidth usage.
Innovation Solution
The implementation of a quantization parameter adaptive mixture-of-experts filter, utilizing a convolution neural network to determine gating weights for expert modules, which combine FIR filters to generate improved image blocks based on quantization parameters and neighboring pixels, enhancing picture quality in video processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional intra prediction and motion compensation techniques are used, then device complexity is reduced, but picture quality deteriorates and compression efficiency is limited
Solution Approach 1:
The patent divides the filtering process into multiple expert modules (e.g., deblocking filter, sample adaptive offset filter, adaptive loop filter) that can be independently selected and applied. Each expert module handles specific types of artifacts or image regions, allowing the system to achieve high picture quality through specialized processing without requiring a single complex filter to handle all cases.
Solution Approach 2:
The patent dynamically adjusts filtering parameters including quantization parameter (QP) adaptive gating weights, filter strength, and filter type selection based on local image characteristics and QP values. This allows the system to optimize picture quality for different compression levels and image regions without increasing overall device complexity through fixed complex processing.
2Loss of energy
If higher compression ratios are achieved through lossy compression, then bandwidth and storage requirements are reduced, but distortion increases and picture quality deteriorates
Solution Approach 1:
The patent implements feedback mechanisms where the encoder and decoder use the same QP adaptive MoE filter processing. The encoder applies filtering to the original image before compression, and the decoder applies identical filtering to the reconstructed image, ensuring that both sides compensate for compression artifacts in the same way. This feedback approach maintains picture quality by systematically reducing distortion introduced by lossy compression.
Solution Approach 2:
The patent converts the harmful effects of quantization and compression artifacts into benefits by using QP as an adaptive parameter to control filter strength. The same quantization process that causes distortion also provides information about the required filtering intensity, allowing the system to apply stronger filtering where compression artifacts are more severe, thereby converting the harm of lossy compression into improved picture quality.
3Manufacturing precision
If QP adaptive gating weights are used to dynamically select expert modules, then picture quality is improved, but computational complexity increases
Solution Approach 1:
The patent uses gating weights to selectively apply filtering only where needed rather than uniformly processing the entire image. The QP adaptive gating mechanism identifies regions with high compression artifacts (typically low-frequency regions with high QP values) and applies filtering primarily to those regions, reducing overall computational complexity while maintaining picture quality in the most critical areas.
Solution Approach 2:
The patent implements dynamic expert module selection where the set of active expert modules and their corresponding gating weights are adjusted based on local QP values and image characteristics. This dynamic approach allows the system to allocate computational resources efficiently, activating more complex filtering only where necessary to improve picture quality while using simpler processing in regions where compression artifacts are minimal.
Data Source
AI summary
Aspects of the disclosure provide methods and apparatuses for video processing. In some examples, an apparatus for video processing includes processing circuitry. The processing circuitry determines gating weights associated with expert modules based at least partially on a quantization parameter (QP) value. The expert modules are configured to process an input block that includes an image block being reconstructed based on the QP value. The processing circuitry then weights the expert modules based on the gating weights to generate an output image block corresponding to the image block.