Last Significant Coefficient Coding With Correlated Coordinate Contexts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data compression techniques, such as those in the H.264/AVC standard, face challenges in efficiently encoding and decoding binary data due to high computational demands and serial processing limitations, particularly in high-quality video decoding where high output speed is required, especially for devices with limited processing power.
Innovation Solution
The proposed solution involves a method for encoding and decoding the last significant coefficient position in a block-based coding scheme by binarizing two-dimensional coordinates of the last significant coefficient, determining contexts for each coordinate, and entropy encoding based on these contexts to improve compression efficiency, while also modifying the context model to exploit correlation between coordinates for better encoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If CABAC (Context Adaptive Binary Arithmetic Coding) is used for entropy coding, then compression efficiency is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the two-dimensional coordinate encoding into two separate one-dimensional coordinate encodings. Each coordinate (x and y) is encoded independently using simplified context models, avoiding the computational complexity of joint two-dimensional context modeling while maintaining compression efficiency.
Solution Approach 2:
The patent transforms the two-dimensional coordinate encoding problem into two separate one-dimensional problems. By treating each coordinate dimension independently and applying separate context models, the patent reduces the complexity from O(N²) to O(N) where N is the coordinate range, while still exploiting spatial correlations through the context model design.
2Productivity
If serial processing is used for entropy coding, then implementation is simpler, but output speed is limited
Solution Approach 1:
The patent divides the encoding task into independent segments (x-coordinate encoding and y-coordinate encoding) that can be processed in parallel. Each coordinate undergoes binarization, context determination, and entropy encoding as separate parallel operations, enabling simultaneous processing without complex interdependencies.
Solution Approach 2:
The patent performs preliminary binarization of coordinates and pre-determination of context models before the actual entropy encoding. This preparation allows the main encoding loop to operate more efficiently with pre-computed contexts, reducing the critical path delay and enabling faster parallel processing.
3Loss of substance
If context modeling is applied to each bin of coordinates, then compression efficiency is improved, but encoding complexity increases
Solution Approach 1:
The patent segments the context modeling into separate models for x-coordinate bins and y-coordinate bins. Each dimension has its own context model that operates independently on its respective bins, reducing the overall complexity from modeling all two-dimensional bins to modeling only one-dimensional bins in each direction.
Solution Approach 2:
The patent reduces the context modeling from two-dimensional space to one-dimensional space by treating each coordinate dimension separately. This dimensional reduction transforms the complexity from O(M×N) context models to O(M) + O(N) context models, where M and N are the number of bins in each dimension.
Data Source
AI summary
Methods and devices are described for entropy coding data using an entropy coder to encode quantized transform domain coefficient data. Last significant coefficient information is signaled in the bitstream using two-dimensional coordinates for the last significant coefficient. The context for bins of one of the coordinates is based, in part, upon the value of the other of the coordinates. In one case, instead of signaling last significant coefficient information, the number of non-zero coefficients is binarized and entropy encoded.