Binarizing Quantization Parameters for Parallel Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional moving picture coding and decoding methods, particularly in the HEVC standard, face challenges in accelerating processing due to dependencies in binarization and arithmetic coding of quantization parameters, leading to reduced throughput and efficiency.
Innovation Solution
The proposed method involves binarizing and decoding quantization parameters without requiring information on maximum values, using bypass arithmetic coding with a fixed probability for the absolute value and sign components, allowing for parallel processing and accelerated processing by eliminating dependencies between sign and absolute value binarization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional binarization methods requiring maximum value information are used, then coding precision is maintained, but processing speed decreases due to sequential dependencies
Solution Approach 1:
The quantization parameter is segmented into two independent parts: a sign component and an absolute value component. The sign component (sign_flag) indicates the sign, while the absolute value component (cu_qp_delta_abs) represents the magnitude. This segmentation allows each component to be binarized and coded independently without requiring maximum value information, eliminating sequential dependencies and enabling parallel processing to improve processing speed.
Solution Approach 2:
The binarization approach is changed from conventional methods that require maximum value information to a new parameter representation that uses sign_flag and cu_qp_delta_abs. This parameter change enables the use of bypass arithmetic coding with fixed probability models, removing the need for context updates and maximum value lookups, thereby accelerating processing while maintaining coding precision.
2Productivity
If conventional arithmetic coding with context updates is used, then coding efficiency is optimized, but processing throughput decreases due to sequential context dependencies
Solution Approach 1:
The context update mechanism is extracted and removed from the arithmetic coding process. Instead of using adaptive context models that require sequential updates based on previously decoded values, the patent applies bypass arithmetic coding with fixed probability models. This extraction eliminates the sequential dependency chain, allowing parallel processing of multiple quantization parameters simultaneously, thereby increasing processing throughput while maintaining acceptable coding efficiency.
Solution Approach 2:
The patent uses simple fixed probability models instead of complex adaptive context models. These fixed models act as 'cheap' computational objects that can be reused without updates, eliminating the need for expensive context maintenance operations. This approach trades some adaptive optimization for significantly improved processing throughput through parallel execution.
3Manufacturing precision
If maximum value information is required for binarization, then accurate quantization is achieved, but processing time increases due to additional lookup operations
Solution Approach 1:
Instead of determining the absolute value first and then applying the sign (the conventional approach requiring maximum value lookups), the patent inverts the process by separately encoding the sign_flag and cu_qp_delta_abs. This inversion allows the absolute value to be represented without reference to maximum bounds, eliminating lookup operations and reducing processing time while maintaining quantization precision through the separate sign component.
Data Source
AI summary
Provided is a moving picture coding method for performing arithmetic coding on a quantization parameter to be used in quantizing a current block that is to be coded and is included in a current picture to be coded, the quantization parameter including a parameter indicating a sign and a parameter indicating an absolute value, the method including: binarizing the parameter indicating the absolute value without requiring information on a maximum value of the absolute value; binarizing the parameter indicating the sign without requiring information on a maximum value of the sign; and performing bypass arithmetic coding, using a fixed probability, sequentially on (i) a latter half of the binarized parameter indicating the absolute value and (ii) the binarized parameter indicating the sign.