Personal identification by image-processing is an important technique for personal reference or verification which can serve a highly valuable service in an information based society. But, it is difficult to identify a person with image processing by computer because a human face is changeable. Although it is difficult, the personal identification from facial images is an attractive subject in computer systems because of its wide application. Therefore, the auhtors propose a method of personal identification from a series of photographic images taken naturally. The method is pattern matching of salient features, constructed automatically from facial images. In this study, in order to approach the realization of a human interface, the main stress falls on automatic salient feature extraction using sequential photographic images taken in normal conditions. At first, from sequential images, eyes coordinates are extracted by detecting eyes winking. Based on these coordinates a full description of the face is defined. Finally, a mosaic is applied to that face to be used for identification. Experimental results show that the proposed method is robust and valid for numerous kinds of facial images in real scenes.
展开▼