Feature Encoding via Major-Unit Selection for Low-Data Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding/decoding technologies struggle to achieve high-efficiency compression and recognition accuracy when considering both human and machine vision simultaneously, particularly in applications like surveillance, intelligent transportation, and smart cities.
Innovation Solution
A method of selecting a major unit from features, converting multi-layer features to single-layer features, and encoding/decoding based on these units, involving channel reduction, resolution adjustment, and importance vector-based selection to reduce data while maintaining performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-layer features are fully encoded to maintain recognition accuracy, then encoding precision is improved, but data volume increases
Solution Approach 1:
The patent extracts only the most important channels from multi-layer features based on importance evaluation, rather than encoding all channels. This selective extraction maintains recognition accuracy while reducing data volume by filtering out less important feature channels.
Solution Approach 2:
The patent applies different encoding strategies to different channels based on their importance. Important channels are encoded with higher precision while less important channels are encoded with lower precision or discarded, achieving local optimization of encoding quality according to each channel's contribution to recognition accuracy.
2Loss of information
If all feature channels are retained to preserve information, then information completeness is improved, but encoding complexity increases
Solution Approach 1:
The patent extracts only the essential feature channels needed for maintaining information completeness, removing redundant channels. This reduces encoding complexity by processing fewer channels while preserving the most important information through selective channel extraction based on importance metrics.
Solution Approach 2:
The patent changes the parameter of channel selection by introducing importance evaluation and dynamically selecting which channels to retain. This parameter-based selection approach simplifies encoding complexity by adapting the number and type of channels processed, rather than uniformly processing all channels.
3Quantity of substance
If channel reduction is performed to decrease data volume, then data compression is improved, but feature quality deteriorates
Solution Approach 1:
The patent applies local quality by differentiating between important and less important channels, applying higher quality preservation to important channels while allowing greater compression for less important channels. This selective quality approach maintains overall feature quality while achieving data volume reduction through differential compression strategies.
Solution Approach 2:
The patent changes the quality parameter adaptively based on channel importance. For important channels, higher quality encoding parameters are used to preserve feature quality, while for less important channels, lower quality parameters are applied to achieve compression, thus balancing data volume reduction with feature quality maintenance.
Data Source
AI summary
A method of encoding a feature according to the present disclosure may include converting a multi-layer feature to a single-layer feature; generating encoding data by transforming the single-layer feature; generating a 2-dimensional (2D) image based on the encoding data; and encoding the 2D image. In this instance, the encoding data may be generated by reducing channels or reducing a resolution of the single-layer feature.