Block-wise Entropy Coding for Neural Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural image compression (NIC) methods face challenges in optimizing rate-distortion performance due to limitations in block-wise entropy coding, which can lead to suboptimal compression efficiency and increased computational resources.
Innovation Solution
A block-wise entropy coding method that partitions images into multiple blocks, compresses each block using a neural network-based encoder, and processes the compressed features through an entropy encoder to generate two compressed bitstreams, improving the rate-distortion performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional hybrid video codec optimizes a single module, then the optimization process is simple, but the overall coding performance does not improve
Solution Approach 1:
The patent divides the image into multiple blocks and applies separate entropy coding to each block. This segmentation allows the system to optimize coding parameters locally for each block rather than using a single global optimization approach, thereby improving overall coding performance while managing complexity through modular processing
Solution Approach 2:
The patent implements block-wise entropy coding where each block can have its own entropy coding parameters and optimization strategy. This local quality approach allows different parts of the image to be coded with different precision and complexity levels, improving overall performance without uniformly increasing system complexity
2Productivity
If block-wise entropy coding is applied to each block, then compression efficiency improves, but computational resources increase
Solution Approach 1:
By segmenting the image into blocks and processing them independently with entropy coding, the patent achieves better compression efficiency through localized optimization. The segmentation enables parallel processing potential and allows the system to focus computational resources on blocks that benefit most from entropy coding
Solution Approach 2:
The patent applies entropy coding selectively to compressed features of each block rather than to the entire image at once. This partial action approach improves compression efficiency by targeting specific blocks that require it, while avoiding the excessive computational cost of applying full entropy coding uniformly across all data
3Manufacturing precision
If compressed features are processed through entropy encoder for each block, then rate-distortion performance is optimized, but memory requirements increase
Solution Approach 1:
The patent processes compressed features in block segments rather than as a single large data structure. This segmentation reduces peak memory requirements by limiting the amount of data that needs to be held in memory at any one time, while still achieving optimized rate-distortion performance through block-wise entropy coding
Solution Approach 2:
The patent extracts and processes only the necessary compressed features for each block through the entropy encoder, rather than processing all possible features. This extraction approach optimizes rate-distortion performance by focusing on relevant information while reducing memory requirements by eliminating unnecessary data storage
Data Source
AI summary
Systems and methods for block-wise entropy coding methods in neural image compression is provided. A method includes: receiving a bitstream that includes an image; partitioning the image into a plurality of blocks; compressing each of the plurality of blocks by a neural network-based encoder; obtaining compressed features by obtaining a compressed feature for each block from among the plurality of blocks in the image; processing the compressed features by an entropy encoder to generate a first compressed bitstream; obtaining a plurality of reshaped compressed features by concatenating the compressed features; processing the plurality of reshaped compressed features by the entropy encoder to generate a second compressed bitstream; and encoding the bitstream including the image based on the second compressed bitstream.