Multi-Frame Face Comparison Using Quality and Diversity Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face-scanning payment systems face challenges in maximizing comparison accuracy due to errors in selecting face image frames based on quality scores, which can result in defects like blur and occlusion, and limited information in single frames.
Innovation Solution
A method and apparatus that select a first face image with the highest quality score and iteratively add face images with maximum average difference in predetermined parameters to a candidate set, ensuring sufficient information for accurate comparison.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only a single face image frame with the highest quality score is selected for comparison, then the comparison process is simple and fast, but the comparison accuracy cannot be maximized due to limited information and potential defects
Solution Approach 1:
The patent segments the face image selection process into multiple stages: first selecting an image with the highest quality score, then iteratively selecting additional images with maximum average difference in predetermined parameters. This segmentation allows the system to balance between simplicity and accuracy by providing optional selection strategies.
Solution Approach 2:
The patent changes the selection parameter from solely quality score to a combination of quality score and average difference in predetermined parameters (such as age, gender, expression, lighting conditions). This parameter change enables the system to select diverse face images that complement each other, thereby maximizing comparison accuracy while managing complexity through configurable parameter weights.
2Measurement precision
If multiple face image frames are selected and compared, then the comparison accuracy is improved through increased information, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-calculating quality scores and predetermined parameters for all candidate face images before the comparison process. This preliminary processing allows the system to quickly identify and select the most suitable images during the actual comparison, reducing real-time processing time while maintaining high accuracy.
Solution Approach 2:
The patent implements partial action by allowing flexible selection of the number of face images to be compared (e.g., selecting 1, 2, or 3 images based on configuration). This partial action approach enables the system to balance between accuracy and processing time by comparing only the necessary number of images rather than processing all available frames, thus reducing time loss while maintaining improved accuracy.
3Measurement precision
If face images with similar characteristics are selected, then the selection process is straightforward, but the diversity of information is limited and comparison accuracy is compromised
Solution Approach 1:
The patent implements feedback by calculating the average difference between selected images and candidate images in predetermined parameters, and using this feedback to guide the iterative selection process. Images with maximum average difference are selected to ensure diversity, while the feedback mechanism prevents selection of redundant images, thereby improving comparison accuracy without excessive complexity.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Implementations of the present specification disclose a comparison method and apparatus based on a plurality of face image frames and an electronic device. The method includes: obtaining a plurality of face image frames of a target object; selecting a first face image with a highest quality score from the plurality of face image frames, and adding the first face image to a candidate image set; in response to that a number of images in the candidate image set is less than a predetermined value, iteratively selecting, from the plurality of face image frames, a second face image that has a maximum average difference from all face images in the candidate image set in a predetermined parameter, to add the second face image to the candidate image set; and comparing the face images in the candidate image set with a stored face image of the target object.