CABAC dQP Binarization Using Separate Absolute Value and Sign
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards, such as HEVC, face inefficiencies in encoding Delta-QP (dQP) due to asymmetric bit allocation and lack of symmetry in probability distribution, leading to suboptimal compression and increased complexity.
Innovation Solution
The Separate Absolute Value and Sign (SAVS) method for binarization in CABAC entropy coding, where dQP is flagged for zero or non-zero values, and if non-zero, its absolute value is mapped using unary codes followed by encoding its sign, or vice versa, to leverage the symmetric distribution of dQP probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional unary coding is used for dQP in CABAC, then the encoding process is simple, but the coding efficiency is suboptimal due to asymmetric bit allocation not matching the symmetric probability distribution of dQP
Solution Approach 1:
The dQP value is segmented into two separate components: the absolute value and the sign. The absolute value is encoded using unary coding, and the sign is encoded separately using a dedicated sign bit. This segmentation allows the encoding to match the symmetric probability distribution of dQP, improving coding efficiency by allocating bits symmetrically around zero rather than using asymmetric unary coding for the entire signed value.
2Loss of information
If separate absolute value and sign encoding is used for dQP, then coding efficiency improves by matching symmetric distribution, but the binarization process complexity increases
Solution Approach 1:
The dQP value is segmented into two separate components: the absolute value and the sign. The absolute value is encoded using unary coding, and the sign is encoded separately using a dedicated sign bit. This segmentation allows the encoding to match the symmetric probability distribution of dQP, improving coding efficiency by allocating bits symmetrically around zero rather than using asymmetric unary coding for the entire signed value.
Solution Approach 2:
Instead of directly encoding the signed dQP value using traditional asymmetric unary coding, the invention inverts the approach by first encoding the absolute value and then separately encoding the sign. This inversion of the encoding sequence and methodology allows the system to exploit the symmetric distribution properties of dQP, achieving better compression while maintaining a relatively simple encoding process.
3Ease of operation
If asymmetric bit allocation is used for dQP encoding, then the encoding structure is simple, but it does not match the symmetric probability distribution of dQP values
Solution Approach 1:
The dQP value is segmented into two separate components: the absolute value and the sign. The absolute value is encoded using unary coding, and the sign is encoded separately using a dedicated sign bit. This segmentation allows the encoding to match the symmetric probability distribution of dQP, improving coding efficiency by allocating bits symmetrically around zero rather than using asymmetric unary coding for the entire signed value.
Solution Approach 2:
Different encoding methods are applied to different parts of the dQP value based on their statistical properties. The absolute value component, which follows a non-negative distribution, is encoded using unary coding. The sign component, which has approximately equal probability for positive and negative values, is encoded using a dedicated sign bit. This local quality approach optimizes the overall encoding efficiency by matching the encoding method to the local statistical characteristics of each component.
Data Source
Figure 1
Figure 2
Figure 3~4B
AI summary
A bit stream, comprising an absolute value of a differential quantization parameter, wherein the differential quantization parameter is a difference between a quantization parameter of a current coding unit and a predicted quantization parameter of the current coding unit and a sign value of the differential quantization parameter, wherein the absolute value and the sign value are separately encoded in an order from the absolute value of the differential quantization parameter to the sign value of the differential quantization parameter and are encoded in the order from the absolute value of the differential quantization parameter to the sign value of the differential quantization parameter in syntax of the bit stream..