Facial identification system based on a high-speed matching algorithm for isodensity lines

In this paper, a high-speed personal identification system which is adaptable for slant faces is presented. In the proposed method, the 25 facial images which are taken with different angles are registered in the database as reference images. The similarity of an input image is first examined between all registered images and the three facial images, which are similar to the input image, are then extracted as the candidate facial images. The isodensity map, which is expected to have the maximum similarity to the input image, is created by the interpolation between two candidate images and the input facial image is judged identical or not with the interpolated images. A template matching is adopted to obtain a similarity between two isodensity lines. In general, template matching requires a lot of time to obtain maximum similarity, while on the otherhand it is very simple to use. To reduce the processing time for the matching efficiently, the techniques which are suited for the line pictures and the Sequential Similarity Detection Algorithm (SSDA) are adopted. Computer simulation using 50 people shows an accuracy of 95.2% correct discrimination. In addition, processing time is reduced to approximately 1/12. From the experimental results, the prospects of using this method are very encouraging.