Image Feature Transmission Using Correlated Matrix Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human vision-based encoding methods for data communication in machine learning applications suffer from low code compression rates, leading to difficulties in ensuring communication quality with increased data volumes, particularly in systems like the Internet of Vehicles and smart cities.
Innovation Solution
A method for transmitting image features using a machine learning model to extract feature matrices, determine correlated matrices, and encode a representative matrix along with its correspondence and feature values, reducing redundancy by analyzing channel correlations through variance homogeneity tests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If human vision-based encoding methods are used for data communication, then the encoding process is simple and intuitive, but the code compression rate is low and communication quality deteriorates with increased data volumes
Solution Approach 1:
The patent replaces traditional human vision-based mechanical encoding methods with a machine learning model that automatically extracts feature matrices from images. This substitution enables the system to identify and encode only the most informative features while eliminating redundant information, thereby achieving high compression rates without sacrificing communication quality.
Solution Approach 2:
The patent extracts essential feature matrices from complete images using a machine learning model. By selecting and encoding only the critical feature components rather than transmitting entire images or all pixel data, the system achieves efficient compression while preserving the most important visual information for downstream tasks.
2Loss of information
If feature matrices from all channels are transmitted without correlation analysis, then complete image information is preserved, but data redundancy increases and communication efficiency decreases
Solution Approach 1:
The patent extracts and transmits only the representative feature matrix from correlated channel groups, eliminating redundant representations. By identifying correlations between channels and selecting only the most informative features for transmission, the system reduces data volume while maintaining complete essential information for accurate image reconstruction and analysis.
Solution Approach 2:
The patent changes the representation parameters by transforming redundant channel data into correlated feature groups. By analyzing correlations and selecting representative features based on information content, the system optimizes the parameter set being transmitted, reducing data volume while preserving the essential information needed for downstream machine learning tasks.
3Loss of information
If correlation analysis is performed on all feature matrices, then redundant information is eliminated, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the feature matrix processing into distinct stages: first extracting feature matrices from images using a pre-trained model, then performing correlation analysis only on extracted features rather than raw pixel data, and finally selecting representative features for transmission. This segmentation reduces computational complexity by operating on compressed feature representations rather than full-resolution images.
Solution Approach 2:
The patent performs preliminary feature extraction using a pre-trained machine learning model before conducting correlation analysis. By pre-processing images to extract meaningful features and storing them in a feature database, the system eliminates the need to perform computationally intensive correlation analysis on raw image data, significantly reducing processing time and complexity during actual transmission operations.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
The invention relates to an image feature transmission method, device and system, and relates to the technical field of communication. The transmission method comprises the following steps: extracting a feature matrix of a to-be-processed image for each channel by using a machine learning model; determining one or more incidence matrix pairs according to a comparison result of the correlation degree between the feature matrixes and a first threshold value; according to the information amount, determining a representation matrix and a represented matrix in the two feature matrixes of each incidence matrix pair; determining a corresponding relation between each representation matrix and each represented matrix; and carrying out quantization processing and coding processing on each representation matrix, the corresponding relation and the maximum characteristic value and the minimum characteristic value in each represented matrix, and then transmitting to a decoding end.