Adaptive Loop Filter Classification for Video Dynamic Range
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding and decoding technologies face challenges in accurately determining the category of adaptive loop filters due to reliance on a maximum dynamic range represented by signal bit width, leading to low video quality and efficiency.
Innovation Solution
The method involves obtaining information for classification, including a target range and intermediate information, and using a classifier group to determine a target category for adaptive loop filters, enhancing accuracy and efficiency by considering pixel information and generated intermediate data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the maximum dynamic range represented by signal bit width is used for classification, then the classification process is simple, but the accuracy of determining adaptive loop filter category is low
Solution Approach 1:
The patent segments the dynamic range into multiple preset ranges (e.g., first dynamic range, second dynamic range, third dynamic range) instead of using a single maximum dynamic range. This segmentation allows the classifier to consider different scale regions of image data, improving classification accuracy while maintaining manageable process complexity through structured range definitions.
Solution Approach 2:
The patent introduces a new dimension of classification by incorporating intermediate information generated during encoding/decoding processes. This adds a temporal or procedural dimension to the spatial dimension of dynamic range analysis, enabling more comprehensive category determination without excessive complexity increase.
2Measurement precision
If more classification information including target range and intermediate information is considered, then the accuracy of adaptive loop filter category determination is improved, but the encoding and decoding complexity increases
Solution Approach 1:
The patent performs preliminary classification actions by pre-defining multiple preset ranges and pre-generating intermediate information during the encoding process. This allows the decoder to use these pre-computed values for efficient classification without performing complex calculations in real-time, thus improving accuracy while maintaining decoding efficiency.
Solution Approach 2:
The patent uses intermediate information generated during encoding as a mediator between the original image data and the final classification decision. This intermediate representation captures essential features while reducing computational complexity, enabling accurate category determination without direct complex analysis of the original data.
3Productivity
If a single maximum dynamic range is used for classification, then the processing is fast, but the video quality is poor
Solution Approach 1:
The patent segments the dynamic range into multiple preset ranges that correspond to different scale regions. This segmentation allows the system to process different parts of the image data with appropriate detail levels, improving video quality while maintaining processing speed through parallel or hierarchical classification approaches.
Solution Approach 2:
The patent applies different classification strategies to different local regions of the image data based on their dynamic range characteristics. This local quality approach allows optimized processing for each region (e.g., using coarser classification for low-dynamic-range regions and finer classification for high-dynamic-range regions), balancing speed and quality.
Data Source
AI summary
In some examples, classification information of image data is obtained, the classification information includes at least one of a target range, intermediate information, or a category subset. A target category is determined according to the classification information, the target category indicates a category of an adaptive loop filter to be used during processing of the image data. The target range includes at least one of a first dynamic range that is determined based on pixel information of the image data and a second dynamic range that is selected from one or more preset ranges, and a preset range has a range width less than a maximum range width for a signal bit width of the image data. The intermediate information is generated during the processing of the image data. The category subset includes at least two subcategories that are respectively determined based on respective classifiers in a classifier group.