Video Compression Parity Hiding Reduces Context-Coded Bins
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as H.264/AVC and developing HEVC, face inefficiencies in encoding and decoding transform coefficients, particularly in the high percentage of bitstream data occupied by quantized transform coefficients, which can be improved by adaptive threshold settings and sign data hiding techniques.
Innovation Solution
The proposed method involves using parity to hide information about the upper left coefficient in a coefficient group, conditional decoding of significant-coefficient and greater-than-one flags based on the parity of the upper left coefficient, and reducing the maximum number of greater-than-one flags to enhance encoding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If quantized transform coefficients are encoded using conventional context modeling, then compression quality is maintained, but the number of context-coded bins increases significantly (occupying 30-80% of bitstream data)
Solution Approach 1:
The patent extracts and removes redundant context-coded bins from the encoding process by identifying that certain coefficient positions (particularly in the upper-left region of transform blocks) do not require separate context modeling. This is achieved by applying adaptive thresholds that eliminate the need for context-coded bins in regions where coefficients are likely to be zero or follow predictable patterns, thereby reducing the overall number of bins without sacrificing compression quality.
Solution Approach 2:
The patent applies different encoding strategies to different regions of the transform block. Adaptive thresholds are applied specifically to the upper-left coefficient groups where redundancy is highest, while other regions maintain conventional encoding. This localized approach reduces the number of context-coded bins in regions where it matters most while preserving compression efficiency in regions where detailed encoding is still necessary.
2Productivity
If adaptive threshold settings are applied to reduce context-coded bins, then compression efficiency improves, but encoding complexity increases
Solution Approach 1:
The patent performs preliminary analysis of the transform coefficient patterns before the main encoding process. By pre-determining which coefficient regions will benefit from adaptive thresholding based on statistical properties and prediction models, the system prepares encoding parameters in advance. This preliminary action allows the main encoding loop to simply apply pre-computed thresholds rather than performing complex real-time analysis, thereby improving compression efficiency without proportionally increasing encoding complexity.
3Device complexity
If the number of greater-than-one flags is reduced, then the number of context-coded bins decreases, but measurement precision of coefficient magnitudes may be affected
Solution Approach 1:
The patent applies greater-than-one flags selectively rather than universally. By using adaptive thresholds, the system applies these flags only in regions where coefficient magnitudes are likely to exceed the threshold, rather than encoding them for all coefficients. This partial application reduces the number of context-coded bins while maintaining precision where it is actually needed, based on statistical analysis of coefficient distributions in different transform block regions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and devices for reconstructing coefficient levels from a bitstream of encoded video data for a coefficient group in a transform unit. Parity hiding is used to signal the parity of the upper left coefficient of the coefficient group. Based on the parity, either the significant coefficient flag or the greater-than-one flag is encoded/decoded for that coefficient, but not both. The greater-than-one flag is encoded/decoded irrespective of whether a maximum number of greater-than-one flags have been encoded/decoded with respect to the other coefficients in the coefficient group.