Mosaic image method: a local and global method

In this paper, a new method to compute eigenimages in principal component a nalysis (PCA) based vision systems is presented. It is called the mosaic im age method. Tn this method, the object is represented as a collection of fe atures and their relative positions (topology). This is a local and global method. Although this method is created to account for the occlusion proble m, it is found that the resulting representation is better than that obtain ed using the traditional optimum representation. A simple algorithm for rec ognition based on the new representation is proposed. Extensive experiments are conducted. More than 110,000 test images with varying degree of occlus ion are used to test the proposed method. It is found that the new method c an accommodate up to 53% occluded parts with a recognition rate of more tha n 95%. To our knowledge, this is the best result in the presence of occlusi on ire PCA-based vision systems. (C) 1999 Pattern Recognition Society. Publ ished by Elsevier Science Ltd. All rights reserved.