Dynamic Entropy Coding With Adaptive Symbol-to-Rank Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dynamic entropy coding techniques are computationally intensive and inefficient when the image statistics do not match the assumed probability distribution, leading to higher compressed file sizes.
Innovation Solution
A two-stage mapping process where symbols are dynamically mapped to ranks, and then ranks are statically mapped to binary codes, using synchronized encoder and decoder look-up tables that update independently without sharing data, to resolve constraint violations and optimize compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional dynamic entropy coding techniques are used to adapt to image statistics, then compression efficiency improves when statistics match assumptions, but computational complexity increases significantly
Solution Approach 1:
The patent segments the encoding process into two distinct stages: (1) a dynamic symbol-to-rank mapping stage that adapts to image statistics, and (2) a static rank-to-binary-code mapping stage that provides computational efficiency. This segmentation allows the system to benefit from adaptive coding without the full computational burden of completely dynamic coding schemes.
Solution Approach 2:
The patent implements dynamic adaptation only in the symbol-to-rank mapping stage, where the encoder dynamically adjusts its mapping based on observed symbol frequencies in the image. The rank-to-binary-code stage remains static, providing a stable computational foundation. This partial dynamics approach balances adaptability with computational efficiency.
2Device complexity
If static mapping of symbols to binary codes is used, then computational complexity is reduced, but compression efficiency deteriorates when image statistics do not match assumed probability distributions
Solution Approach 1:
The patent introduces dynamic adaptation in the symbol-to-rank mapping stage, allowing the encoder to adjust its mapping based on actual image statistics observed during encoding. This dynamic component enables the system to adapt to various image types and statistics, improving compression efficiency without requiring completely dynamic binary code assignment.
Solution Approach 2:
The patent introduces an intermediate 'rank' concept that acts as a mediator between symbols and binary codes. Instead of directly mapping symbols to binary codes, the system first maps symbols to ranks based on their frequency, then maps ranks to binary codes. This intermediary layer allows the system to adapt to image statistics while maintaining a stable binary code structure.
3Productivity
If dynamic entropy coding is implemented to optimize compression for natural images, then compression performance improves, but the system requires prior knowledge of probability distributions which increases system complexity
Solution Approach 1:
The patent performs preliminary dynamic symbol-to-rank mapping based on observed symbol frequencies before the final rank-to-binary-code conversion. This preliminary adaptation allows the system to prepare optimized mappings based on actual image statistics without requiring pre-computed probability distributions, reducing system complexity while maintaining compression performance.
Solution Approach 2:
The encoder performs self-adaptation by observing symbol frequencies during the encoding process and automatically adjusting its symbol-to-rank mapping accordingly. This self-service capability eliminates the need for external probability distribution information or complex pre-processing, allowing the system to optimize compression performance autonomously.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and a method are disclosed for encoding data for transmission from a source to a destination over a communication channel, including obtaining a symbol stream comprising a plurality of symbols, determining a rank of a first obtained symbol of the plurality of symbols from a first encoder look-up table, LUT, encoding, at an encoder, the rank of the first obtained symbol, generating a new frequency entry for the first obtained symbol by incrementing an initial histogram frequency entry of the first obtained symbol after encoding the rank of the first obtained symbol, determining, based on the new frequency entry of the first obtained symbol, whether the rank of the first obtained symbol of the plurality of symbols has a constraint violation with a rank of a first violating symbol in the first encoder LUT, swapping, upon determining that the constraint violation is present, the rank of the first obtained symbol and the rank of the first violating symbol in the first encoder LUT so the constraint violation is resolved, and generating a compressed bit-stream by iteratively applying an encoding function to each symbol of the plurality of symbols.