Entropy Coding Probability Updates Without Lookup Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding and decoding technologies face limitations in adapting symbol probabilities efficiently, particularly due to high memory usage, coarse adaptation rates, and issues with probabilities approaching 0 or 1, which lead to increased encoding costs and limited flexibility in handling varying symbol statistics.
Innovation Solution
The implementation of an explicit update function for symbol probabilities in entropy encoders and decoders, which is truncated at a threshold value and has a varying rate of adaptation, avoiding the use of look-up tables and allowing for finer adaptation rates, is proposed. This function can take the form of linear or exponential updates, optimized for integer arithmetic to prevent extreme probability values and improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If look-up tables are used for probability updates in arithmetic coding, then adaptation to video content is achieved, but memory usage increases and the number of probabilities is limited
Solution Approach 1:
The patent extracts the probability update logic from fixed look-up tables and implements it through explicit mathematical functions. The update function pn = pn-1 - (pn-1 - delta) >> m allows the system to compute probability updates on-the-fly without storing extensive lookup tables, thereby reducing memory requirements while maintaining adaptation capability.
Solution Approach 2:
The patent changes the parameter representation by using an explicit update function with controllable parameters (delta and m) instead of fixed table entries. This allows dynamic adjustment of adaptation characteristics through parameter modification rather than requiring separate table structures for different probability states.
2Productivity
If explicit update functions are used with integer arithmetic, then implementation efficiency is improved, but the degree of adaptation is very limited with coarse adaptation rates
Solution Approach 1:
The patent introduces dynamic adaptability by allowing the parameters delta and m to vary based on the coding context and symbol frequency. The adaptation rate becomes variable rather than fixed, enabling fine-grained control over probability updates while maintaining integer arithmetic efficiency. The update function can adapt its behavior dynamically based on the current state and input characteristics.
3Productivity
If probability values are allowed to approach 0 or 1, then coding efficiency for common symbols is improved, but encoding cost increases when the low-probability symbol actually appears
Solution Approach 1:
The patent applies beforehand cushioning by maintaining a minimum probability threshold (delta) that prevents probabilities from approaching zero too closely. The update function pn = pn-1 - (pn-1 - delta) >> m ensures that even when a symbol appears frequently, its probability remains bounded away from zero by at least delta, providing a safety buffer that prevents excessive encoding costs for rare symbol occurrences.
Data Source
AI summary
Methods and apparatus are provided for explicit updates for symbol probabilities of an entropy encoder or decoder. An apparatus includes a video encoder having an entropy encoder for encoding symbols for picture data for at least a portion of a picture. An explicit update function is used to update probabilities of the symbols processed by the entropy encoder. The explicit update function at least one of is truncated at a threshold value and has a varying rate of adaptation.