Facial Recognition Feature Extraction Using Hierarchical Example Pyramid
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing security image archives face challenges in rapidly searching through numerous images to detect incidents or identify criminals due to degraded facial recognition performance caused by changes in posture, lighting, or expression.
Innovation Solution
A method of extracting image features using a pyramid structure with hierarchical levels, generating codebooks from stored images, and calculating similarity between input images and codebooks to enhance facial recognition, involving clustering, vector quantization, and sparse coding to improve feature extraction and recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional facial recognition methods are used on stored images, then the system is simple to implement, but recognition accuracy degrades due to changes in posture, lighting, or expression
Solution Approach 1:
The patent segments the facial recognition process into multiple hierarchical levels (example pyramid with multiple levels). Each level processes images at different resolutions and feature granularities, allowing the system to capture both global facial structure and local details, thereby improving recognition accuracy under varying conditions while maintaining manageable complexity through modular processing
Solution Approach 2:
The patent introduces a hierarchical dimension to the feature extraction process by constructing an example pyramid with multiple levels. This multi-level hierarchical structure enables the system to analyze images at different scales and abstractions, transforming the single-level recognition problem into a multi-dimensional feature space that better handles variations in posture, lighting, and expression
2Quantity of substance
If security image archives store numerous images from multiple cameras, then the archive capacity increases, but the time required to search and detect incidents increases
Solution Approach 1:
The patent applies preliminary action by pre-processing images and pre-extracting features at multiple hierarchical levels during the example pyramid construction phase. This preliminary feature extraction and organization into hierarchical levels enables rapid similarity comparison during incident detection, significantly reducing search time in large archives without sacrificing comprehensive image storage capacity
Solution Approach 2:
The patent creates multiple representations (copies) of facial features at different hierarchical levels within the example pyramid. Instead of searching through all original images directly, the system compares query images against these pre-computed hierarchical feature copies, enabling efficient search through large archives by operating on compressed feature representations rather than full-resolution images
Data Source
AI summary
At least one example embodiment discloses a method of extracting a feature of an input image. The method includes constructing an example pyramid including at least one hierarchical level based on stored example images, generating a codebook in each of the at least one hierarchical level, calculating a similarity between the codebook and the input image, and extracting a feature of the input image based on the similarity.