AI Image Encoding With Log-Domain Variance Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI image compression technologies require significant computational resources for variance adjustment, leading to inefficiencies in the encoding and decoding processes.
Innovation Solution
Convert the multiplication operation between variance and gain into an addition operation between the index of variance and gain converted to a logarithm domain, and utilize a probability distribution table based on quantized indices for entropy encoding and decoding, reducing computational overheads while maintaining compression performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiplication operation between variance and gain is used for variance adjustment, then compression performance is maintained, but computation amount increases and efficiency decreases
Solution Approach 1:
The patent transforms the variance adjustment operation from multiplication in linear domain to addition in logarithm domain by changing the parameter representation. Specifically, it uses log(variance) + log(gain) instead of variance × gain, which maintains the same mathematical relationship while enabling more efficient computation through lookup tables and bitwise operations.
Solution Approach 2:
The patent replaces the traditional multiplication operation with an addition operation in the logarithm domain. This substitution fundamentally changes the computational mechanism from complex multiplication to simpler addition, which can be implemented more efficiently using lookup tables and bitwise shift operations instead of traditional arithmetic units.
2Measurement precision
If conventional entropy encoding with variance adjustment is used, then compression accuracy is maintained, but bit rate overheads increase
Solution Approach 1:
The patent extracts only the essential information needed for entropy encoding by representing variance through indices in the logarithm domain. Instead of encoding the full variance value, it encodes compact index representations that can be efficiently mapped to probability distribution tables, reducing the bit rate overhead while preserving the necessary statistical information for accurate compression.
3Reliability
If variance adjustment with multiplication operation is performed, then compression performance is ensured, but computational complexity increases
Solution Approach 1:
The patent performs preliminary preparation by creating lookup tables that store pre-computed logarithm values and probability distribution parameters. During actual encoding/decoding operations, the system simply retrieves and combines these pre-prepared values through addition and bitwise operations, avoiding the need for complex real-time multiplication calculations and reducing overall computational complexity.
Data Source
AI summary
Embodiments of this application provide an encoding method, a decoding method, and an electronic device. The encoding method includes: obtaining a to-be-encoded image; performing feature extraction on the to-be-encoded image to obtain a first feature map; determining a probability distribution parameter corresponding to the first feature map, where the probability distribution parameter includes a first index of a variance; performing first adjustment on the first feature map based on a first gain vector to obtain a second feature map; performing second adjustment on the first index based on a second gain vector to obtain a second index, where the second gain vector is obtained by converting the first gain vector to a logarithm domain, and the second adjustment is an addition operation; and performing entropy encoding on the second feature map based on the second index to obtain a bitstream.