METHOD AND DEVICE FOR COLLATING FACE IMAGE

摘要

PROBLEM TO BE SOLVED: To enable a high-reliability collation by identifying a person by finding reliability showing how much the similarity found from two face images is reliable. SOLUTION: A person face image A obtained from a video camera 101 is stored in an image memory 103 through I/F 115 and a face image B obtained by an image scanner 102 from the ID card photograph of the person is stored in an image memory 104 through I/F 116. Feature vectors are found from the stored images A and B and their similarity is calculated. This similarity is calculated between the face images A and B and sample images read out of a sample image data base 110 and the reliability of the similarity between the face images A and B is calculated as reliability from the number of sample images whose similarity is smaller than a predetermined threshold. The reliability and similarity are compared with predetermined thresholds and when they are larger, it is judged that the person is an identical person, thereby outputting the result from an output terminal 14. When they are smaller than the thresholds, on the other hand, the face images A and B are displayed on a monitor 111.