Long-Range Face Recognition Best-Shot Selection by Image Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional long-range face recognition technologies face challenges due to high computing resource demands and low search performance caused by inconsistent image quality, as they lack effective measurement of quality elements and criteria tailored to the object and environment, leading to degraded recognition performance.
Innovation Solution
A method and apparatus that analyze face images for resolution, focus, and illumination quality, selecting a best shot image based on predefined criteria, and transmit it for face recognition, optimizing the face recognition system's performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If face images are immediately transmitted without quality measurement, then search performance is provided quickly, but recognition performance degrades due to low image quality
Solution Approach 1:
The patent applies preliminary action by measuring quality elements (resolution, focus, pose, illumination) of detected face images before transmission to the face recognition system. This preliminary quality assessment ensures that only images meeting predetermined criteria are transmitted, preventing degradation of recognition performance while maintaining efficient search response time.
2Reliability
If quality measurement and best shot selection are implemented, then recognition performance improves, but computing resources and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the quality measurement process into four independent modules: resolution analysis, focus analysis, pose analysis, and illumination analysis. Each module evaluates a specific quality element separately, making the overall system more manageable and efficient. This segmented approach allows the system to assess image quality comprehensively without excessive computational complexity.
Solution Approach 2:
The patent applies parameter changes by establishing predetermined reference values for each quality element (resolution, focus, pose, illumination) and using these thresholds to filter images. By changing the evaluation parameters into quantifiable reference values, the system can efficiently determine whether images meet quality criteria without complex subjective assessment, thus improving recognition performance while controlling system complexity.
3Device complexity
If limited quality criteria are used, then processing is simplified, but image quality measurement becomes insufficient for diverse environments
Solution Approach 1:
The patent applies universality by designing quality measurement criteria that are universally applicable across different environments and object types. The four quality elements (resolution, focus, pose, illumination) serve as universal indicators that can assess image quality regardless of specific environmental conditions or the particular object being detected, thereby maintaining simplicity while achieving environmental adaptability.
Data Source
AI summary
A method and apparatus for detecting the best shot in a long-range face recognition system is provided. The method of detecting the best shot includes detecting a facial area in an image received from the outside, calculating a quality element measurement value of the face image by analyzing the facial area, and selecting a best shot face image, among the detected facial areas, based on the quality element measurement value.