Average Face Image Subspace Recognition for Verification Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image recognition technologies face inefficiencies in setting sample points for comparing image features, leading to increased workload and memory requirements, especially when dealing with a large number of registered images, and result in prolonged verification times during 1:N matching processes.
Innovation Solution
An image recognition apparatus that creates an average image from a group of images, detects sample points on both the input and registered images based on subspaces generated from the average image, and calculates similarity between feature vectors to efficiently verify identities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sample points are set in advance for all registered images, then local feature comparison capability is improved, but workload increases with the number of registered images
Solution Approach 1:
The patent applies preliminary action by creating an average image from the image group in advance and detecting sample points on this average image before verification. This pre-detection of sample points on the average image eliminates the need to set sample points for each individual registered image, thus maintaining local feature comparison capability while reducing workload that would otherwise increase with the number of registered images.
Solution Approach 2:
The patent uses the average image as a universal reference that represents the entire image group. The sample points detected on the average image can be applied to compare against any number of registered images, making the sample point set universal rather than image-specific. This allows the system to maintain high measurement precision across all comparisons without increasing workload proportionally to the number of registered images.
2Measurement precision
If sample points are stored on memory for all registered images, then identification accuracy is improved, but memory capacity requirements increase
Solution Approach 1:
The patent merges multiple registered images into a single average image, combining the information from all images in the group. By detecting sample points on this consolidated average image rather than storing separate sample point data for each registered image, the system maintains identification accuracy while dramatically reducing memory capacity requirements from O(N) to O(1) where N is the number of registered images.
3Productivity
If conventional fixed pixel verification is used, then processing speed is maintained, but distinction capability deteriorates with image distortion and deviation
Solution Approach 1:
The patent changes the reference parameter from fixed pixels in individual registered images to dynamically detected sample points on an average image. This parameter change allows the sample points to adapt to image distortion and deviation while maintaining processing speed, because the sample point detection is performed once on the average image and then applied consistently across all verifications.
Data Source
AI summary
An average face-image is created from images belonging to a face-image group and sample points are specified on the average face-image. Sample points are detected on images in the face-image group corresponding to sample points on the average face-image thereby creating recognition subspaces of feature vectors. Sample points are detected on registered images and an input image corresponding to the sample points on the average face-image by using the recognition subspaces. A degree of similarity between the feature vectors in the recognition subspaces of the registered images and the input image is calculated at the detected sample points. Finally, a person is verified based on the degree of similarity.