Adaptive Binarization in Arithmetic Coding for Changing Symbol Probabilities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video coding standards, such as H.265/HEVC and VP8/VP9, have fixed binarization schemes that do not adapt to changing probability distributions, leading to suboptimal compression efficiency due to the use of pre-defined binarizers tied to context modeling, which can result in longer binary sequences and reduced compression performance.
Innovation Solution
Implementing adaptive binarizer selection methods that allow for the use of modified binarizers based on estimated or historical probability distributions, enabling the selection or construction of binarizers that better match the actual coding distributions, thereby reducing the length of binary sequences and improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed pre-defined binarizer is used as prescribed by the standard, then the context modeling is simplified and the implementation is easier, but the binary sequence length increases and compression efficiency deteriorates
Solution Approach 1:
The patent applies dynamics by transitioning from a static, fixed binarizer prescribed by standards to a dynamic, adaptive binarizer that adjusts its parameters based on the actual probability distribution of the data being encoded. The binarizer is updated during the encoding process to match the statistical characteristics of the input data, thereby optimizing compression efficiency while maintaining implementation feasibility through controlled adaptation mechanisms.
Solution Approach 2:
The patent implements parameter changes by modifying the binarizer's internal parameters (such as threshold values and probability estimates) based on the observed data distribution. Instead of using fixed parameters defined by the standard, the system dynamically adjusts these parameters to reflect the actual statistical properties of the encoded data, achieving better compression ratios without fundamentally changing the overall encoding architecture.
2Loss of information
If a modified binarizer is selected based on probability distribution estimation, then the binary sequence length is reduced and compression efficiency is improved, but the device complexity and computational overhead increase
Solution Approach 1:
The patent applies self-service by enabling the binarizer to automatically adapt to the data distribution without requiring external intervention or complex configuration. The system performs its own probability distribution estimation and adjusts its binarization parameters autonomously during the encoding process, reducing the need for additional complex control mechanisms while achieving improved compression efficiency.
Solution Approach 2:
The patent implements feedback by using the observed frequency of symbols during encoding to continuously update the binarizer's probability estimates. This feedback loop allows the system to refine its binarization strategy based on actual data characteristics, improving compression efficiency while keeping the complexity manageable through iterative adaptation rather than exhaustive optimization.
3Loss of information
If the binarizer is updated adaptively during encoding, then the compression performance is optimized, but the encoding time and processing speed increase
Solution Approach 1:
The patent applies partial action by updating only the necessary parameters of the binarizer based on the current data distribution rather than completely re-optimizing the entire binarization scheme. This selective adaptation approach allows the system to improve compression performance by focusing computational resources on the most impactful parameter adjustments, thereby minimizing the impact on encoding speed.
Solution Approach 2:
The patent implements periodic action by updating the binarizer parameters at regular intervals or based on predefined triggers (such as after processing a certain number of symbols or when significant distribution changes are detected). This periodic update strategy balances the need for optimization with the requirement for maintaining encoding speed, avoiding continuous parameter adjustments that would excessively slow down the encoding process.
Data Source
Figure 1A~1B
Figure 2A~2B
Figure 3A~3B
AI summary
Methods and devices for image and video coding using arithmetic coding. The binarization of symbols in an encoder and decoder is adaptive based on changes to the probability distribution as symbols are encoded/decoded. A binarizer may be generated based upon a probability distribution, used to binarize a symbol, and then the probability distribution is updated based on the symbol. Updates to the binarizer may be made after each symbol, after a threshold number of symbols, or once the updated probability distribution differs by more than a threshold amount from the probability distribution used in generating the current binarizer. The probability distributions may be context-specific.