Transform Coefficient Context Modeling for Faster Entropy Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in optimizing memory access and computational efficiency during entropy coding of transform coefficients, particularly due to non-contiguous memory access and the need for numerous context models, which hampers decoder speed and overall compression performance.
Innovation Solution
The proposed solution involves arranging transform coefficient values along anti-diagonal lines in line buffers and using a scan order that optimizes memory access, reducing non-contiguous memory access and enabling parallel computing, thereby improving computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If transform coefficients are coded using conventional scan orders, then coding coverage is achieved, but memory access becomes non-contiguous reducing decoder speed
Solution Approach 1:
The patent inverts the conventional approach by arranging transform coefficients in reverse zigzag order and using corresponding reversed scan patterns for context modeling. This inversion transforms non-contiguous memory access into contiguous access, enabling parallel computing and significantly improving decoder speed while maintaining coding effectiveness.
2Manufacturing precision
If numerous context models are used for transform coefficient coding, then coding precision is improved, but computational overhead increases reducing efficiency
Solution Approach 1:
The patent segments the transform coefficient block into multiple sub-blocks and applies different scan orders and context models to each sub-block. This segmentation allows the use of fewer context models per sub-block while maintaining overall coding precision, thereby reducing computational overhead and improving processing efficiency.
Solution Approach 2:
The patent applies context modeling selectively to only those transform coefficients that require it based on their position and characteristics, rather than uniformly applying numerous context models to all coefficients. This partial action approach maintains coding precision for critical coefficients while reducing unnecessary computational operations.
3Adaptability or versatility
If non-contiguous memory access is used for transform coefficients, then coding flexibility is maintained, but parallel computing capability is reduced
Solution Approach 1:
The patent inverts the memory arrangement by storing transform coefficients in reverse zigzag order in contigous memory locations, which enables efficient parallel computing while maintaining the flexibility to apply different scan orders for context modeling. This inverted arrangement allows parallel processing units to access multiple coefficients simultaneously without compromising coding adaptability.
Data Source
AI summary
An encoded bitstream is decodable by a processor configured to execute instructions to store, in a first line buffer, first values of a first scan-order diagonal line scanned immediately before a current scan-order diagonal line of a transform block; and store, in a second line buffer, second values of a second scan-order diagonal line scanned immediately before the first scan-order diagonal line. The first values of the first line buffer and the second values of the second line buffer are interleaved in a destination buffer. Using the destination buffer, a probability distribution is selected for coding a current value of the current scan-order diagonal line. The current value is entropy decoded from the bitstream using the probability distribution. One of the second line buffer or the first line buffer is replaced with current values of the current scan-order diagonal line for coding values of an immediately subsequent scan-order diagonal line.