Image Decoding Quantization Parameter Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image coding and decoding methods face challenges in efficiently processing high-resolution images, particularly in determining suitable block division structures and quantization processes, which affect coding efficiency and computational load, especially for ultra-high resolution images like 360 VR content.
Innovation Solution
The method involves parsing division depth and direction information of blocks, obtaining prediction quantization parameters from neighboring units, calculating a difference quantization parameter, and using it to determine an appropriate quantization parameter for efficient decoding, allowing for adaptive quantization and enhanced coding efficiency through complex tree structures like quad, binary, and ternary trees.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If quantization is performed in a great block unit to reduce data amount and improve coding efficiency, then compression ratio increases, but apparatus performance degrades from the viewpoint of rate control and picture quality
Solution Approach 1:
The patent divides the image into multiple blocks of different sizes (e.g., 8x8, 16x16, 32x32 pixels) and applies different quantization parameters to each block. This segmentation allows the system to balance coding efficiency and picture quality by using finer granularity where needed while maintaining coarser blocks where compression is prioritized, thus resolving the contradiction between coding efficiency and rate control performance
Solution Approach 2:
The patent applies different quantization parameters (QP values) to different blocks based on their characteristics and position in the image. Important regions receive higher QP values for better quality, while less important regions use lower QP values for better compression. This local differentiation resolves the contradiction by allowing great block units in some areas while maintaining fine granularity in others
2Productivity
If intra prediction and inter prediction are further subdivided to increase compression efficiency, then residual signal data amount decreases, but computational load of video processing greatly increases
Solution Approach 1:
The patent dynamically adjusts the prediction subdivision level based on image characteristics and coding requirements. Instead of always using the maximum number of prediction modes, the system adapts the prediction process to achieve sufficient compression efficiency while minimizing computational load, resolving the contradiction between compression efficiency and processing complexity
Solution Approach 2:
The patent changes the quantization parameter (QP) values dynamically during the encoding process based on block characteristics, prediction mode, and coding efficiency considerations. By adjusting QP parameters rather than always using fixed high-subdivision prediction, the system achieves compression efficiency without proportionally increasing computational load
3Ease of manufacture
If a conventional block division method is used for image coding, then the process is simple, but it is not suitable for coding of high resolution images and ultra-high resolution images
Solution Approach 1:
The patent extends conventional block division by introducing multiple block size categories (8x8, 16x16, 32x32) and allowing flexible combination of these blocks to represent any image resolution. This segmented approach maintains the simplicity of conventional blocking while adapting to high resolution images through hierarchical composition, resolving the contradiction between simplicity and adaptability
Solution Approach 2:
The patent adds an additional dimension to block division by introducing variable block sizes beyond the conventional fixed sizes. This dimensional extension allows the same blocking framework to handle everything from standard definition to ultra-high resolution images, enhancing adaptability while preserving the simplicity of the conventional approach
Data Source
AI summary
A method for image decoding includes: a step of parsing division depth information and division direction information of a block to be decoded; a step of acquiring division structure information, of a unit to be decoded in the block, corresponding to the parsed division depth information and division direction information; a step of acquiring, based on the division structure information, one or more prediction quantization parameters from at least one neighboring unit corresponding to the unit to be decoded; a step of acquiring a differential quantization parameter for the unit to be decoded; a step of acquiring a quantization parameter from the differential quantization parameter and the one or more prediction quantization parameters; and a step of performing decoding corresponding to the unit to be decoded, by using the acquired quantization parameter.