Image Compression via Spatial Channel Redundancy Grouping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image compression methods fail to effectively utilize spatial and channel redundancy features to enhance compression encoding rates, leading to inefficient storage and transmission of image data.
Innovation Solution
The method involves acquiring a target image, performing feature extraction to obtain a feature map, grouping channels to form second feature maps, and applying spatial and channel context feature extractions to determine redundancy features, which are then used to generate compression information for deep compression processing, resulting in improved encoding rates through both spatial and channel redundancy compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If traditional image compression methods are used, then the compression process is simple, but the compression encoding rate is low and the compressed file size is large
Solution Approach 1:
The patent segments the feature map into multiple channels and processes them separately through channel splitting. This allows independent compression of different feature channels, enabling more sophisticated compression strategies for each channel while maintaining overall efficiency. The segmentation of compression into spatial and channel contexts further applies this principle to handle different types of redundancy separately.
Solution Approach 2:
The patent introduces a channel dimension for feature organization, transforming the traditional 2D spatial processing into 3D channel-spatial processing. This additional dimension enables channel-wise compression operations and allows the model to exploit channel correlations for improved compression ratios without significantly increasing computational complexity.
2Productivity
If spatial and channel context feature extraction is performed, then the compression encoding rate is improved, but the computational complexity increases
Solution Approach 1:
The patent performs feature extraction and channel splitting before the main compression process. By preparing the feature map in advance with organized channels and extracted spatial contexts, the subsequent compression operations can proceed more efficiently. This preliminary organization reduces the computational burden during the actual compression phase.
Solution Approach 2:
The patent extracts and removes redundant spatial and channel context information from the feature map before final compression. By identifying and eliminating these redundancies in advance, the remaining data requires fewer bits for representation, improving the compression encoding rate without requiring excessive computational resources during decoding.
3Loss of substance
If feature map channels are grouped and compressed separately, then the redundancy compression is enhanced, but the processing time increases
Solution Approach 1:
The patent divides the feature map channels into multiple groups or splits them individually, allowing parallel processing of different channel segments. This segmentation enables the compression algorithm to work on smaller subsets simultaneously, reducing the overall processing time while still achieving comprehensive redundancy compression across all channels.
Solution Approach 2:
The patent combines spatial context information with channel-wise compression results in a unified compression framework. By merging these two compression approaches, the system achieves enhanced redundancy removal without requiring separate complete processing passes, thereby reducing total processing time while maintaining high compression ratios.
Data Source
AI summary
An image compression method comprises: performing feature extraction on the target image to obtain a first feature map comprising a plurality of channels; grouping the channels of the first feature map to obtain a plurality of second feature maps; performing spatial context feature extraction on the second feature maps to determine first spatial redundancy features corresponding to the second feature maps; and performing channel context feature extraction on the second feature maps to determine first channel redundancy features corresponding to the second feature maps; determining compression information corresponding to each of the second feature maps based on a first spatial redundancy feature and a first channel redundancy feature corresponding to each of the second feature maps and thus determining first compressed data corresponding to the target image, and performing deep compression processing based on the first feature map to determine second compressed data corresponding to the target image.