Significant Coefficient Coding With Adaptive CABAC Contexts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video data compression and decompression systems face inefficiencies in entropy encoding, particularly with existing CABAC techniques, which do not fully optimize bit rate and image quality due to limitations in context modeling and bit stream termination methods.
Innovation Solution
The implementation of advanced CABAC encoding techniques with adaptive context modeling and improved bit stream termination methods, including the use of fixed bits schemes and context variable adjustments, to enhance encoding efficiency and bit stream alignment, allowing for better compression and decompression of video data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional CABAC entropy encoding is used, then encoding process is simple, but encoding efficiency and bit rate optimization are insufficient
Solution Approach 1:
The patent implements dynamic context modeling where context variables are adaptively adjusted during encoding based on previously encoded significant coefficients. The context state transitions between different states (e.g., STATE_A, STATE_B, STATE_C) depending on the values of significant coefficients, allowing the encoding process to dynamically adapt to local characteristics of the quantized transform coefficients, thereby improving encoding efficiency without requiring overly complex static models
Solution Approach 2:
The patent changes encoding parameters (context variables) based on the values of significant coefficients. Specifically, different context variables are selected or adjusted according to whether coefficients are greater than 1, equal to 1, or less than 1, allowing the encoding strategy to adapt to the magnitude distribution of coefficients and optimize bit rate efficiency for different coefficient patterns
2Productivity
If standard bit stream termination is used, then processing is straightforward, but bit stream alignment and efficiency are suboptimal
Solution Approach 1:
The patent performs preliminary actions during the encoding process by maintaining context state information and tracking the encoding progress through different states. The context variables are prepared and adjusted in advance based on significant coefficient patterns, so that when bit stream termination is needed, the encoder is already in an optimized state that facilitates efficient termination without requiring complex post-processing or real-time calculations at the termination point
3Manufacturing precision
If existing CABAC techniques are applied, then implementation is straightforward, but image quality and bit rate optimization are limited
Solution Approach 1:
The patent applies local quality optimization by treating different regions of the quantized transform coefficient block with different encoding strategies. Based on the local characteristics of significant coefficients (their positions, magnitudes, and patterns), the context variables are locally adapted to match the local statistical properties. This allows the encoder to achieve better image quality in regions with important visual information while using more compact encoding in regions with less significant data, thereby optimizing the overall bit rate efficiency
Data Source
Figure 1~4
Figure 5
Figure 6
AI summary
A data encoding method for encoding an array of data values as data sets and escape codes for values not encoded by the data sets, an escape code comprising a unary coded portion and a non-unary coded portion comprises the steps of: setting a coding parameter defining a minimum number of bits of a non-unary coded portion, the coding parameter being between 0 and a predetermined upper limit; adding an offset value of 1 or more to the coding parameter so as to define a minimum least significant data portion size; generating one or more data sets indicative of positions, relative to the array of data values, of data values of predetermined magnitude ranges, so as to encode the value of at least one least significant bit of each data value; generating, from at least the part of each data value not encoded by the one or more data sets, respective complementary most-significant data portions and least-significant data portions, such that the most-significant data portion of a value represents zero or more most significant bits of that portion, and the respective least-significant data portion is dependent upon a number of least significant bits of that part, the number of least significant bits being greater than or equal to the minimum least significant data portion size; encoding the data sets to an output data stream; encoding the most significant data portions to the output data stream; and encoding the least-significant portions to the output data stream.