Adaptive QP Decoding for Lower Video Quantization Distortion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large quantization distortion in picture encoding and decoding processes during video compression.
Innovation Solution
A method and apparatus for video decoding and encoding that involves analyzing a code stream to determine quantization parameter (QP) values for coding units (CUs) and quantization groups (QGs) within a picture frame, using various division rules and adaptive QP adjustments based on surrounding pixel information to reduce distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If uniform quantization parameter is used for entire coding unit, then encoding complexity is reduced, but quantization distortion increases
Solution Approach 1:
The coding unit is divided into multiple quantization groups (QGs), where each QG can have its own independently determined quantization parameter. This segmentation allows different regions of the image to be quantized with appropriate precision levels, reducing overall quantization distortion while maintaining manageable encoding complexity through structured division.
Solution Approach 2:
Different quantization parameters are assigned to different quantization groups based on local image characteristics such as flatness, texture, and frequency content. This local quality approach ensures that regions requiring higher precision (low distortion) receive appropriate QP values, while simpler regions use lower precision, balancing overall distortion reduction with encoding complexity.
2Manufacturing precision
If quantization parameter is adjusted for each pixel, then quantization distortion is minimized, but processing time increases
Solution Approach 1:
Instead of processing each pixel individually, the image is segmented into quantization groups that aggregate multiple pixels into manageable units. This segmentation reduces the number of independent QP determination operations while still providing fine-grained control over quantization precision, thereby minimizing processing time compared to pixel-level processing.
Solution Approach 2:
The patent applies partial quantization parameter adjustment at the QG level rather than full pixel-level adjustment. This partial action approach provides sufficient distortion reduction for most applications while significantly reducing the computational overhead associated with complete pixel-level optimization.
3Manufacturing precision
If coding unit is divided into multiple quantization groups, then quantization distortion is reduced, but device complexity increases
Solution Approach 1:
The coding unit is divided into quantization groups with standardized division rules that balance precision improvement with complexity control. The segmentation is designed to be systematic and rule-based, allowing the complexity increase to be managed through predictable structural patterns rather than arbitrary divisions.
Solution Approach 2:
The patent utilizes parameter changes in the form of different QP values assigned to different QGs based on image characteristics such as flatness and texture. These parameter changes are applied in a controlled manner through derived QP mechanisms that adjust precision levels without proportionally increasing device complexity.
Data Source
AI summary
A decoding method includes: analyzing a code stream to obtain one or more picture frames, wherein for each of the one or more picture frames, the picture frame comprises one or more coding units (CUs); determining a plurality of quantization parameter (QP) values for the picture frame, wherein a CU of the one or more CUs comprises a plurality of pixels, each of the plurality of pixels corresponds to a respective QP value in the plurality of QP values, and at least two pixels of the plurality of pixels have different QP values, and for each of the plurality of pixels, the respective QP value corresponding to the pixel is adaptively adjusted based on information of one or more reconstructed pixels surrounding the pixel; and decoding, based on the plurality of QP values, the picture frame.