Automatic Human Face Recognition from Still Group Images Based on Image Processing Techniques

Abstract

The most fundamental and crucial technique in face recognition is face matching. Although many different algorithms have been used in various types of research in the field of human face detection, the problem of face detection and recognition persists because it is difficult to achieve reliable face matching under a variety of shooting conditions such as lighting changes, face position differences, or viewing angle differences. In this research, two-dimensional Normalized cross-correlation NCC was used to match the image of suspects with the captured image. The designed recognition algorithm was implemented to recognize a target face in an image of several faces (a Group image). Several different cases were tested. In group image #1, when the target image existed, a high normalized cross-correlation (NCC) value of 0.9307 was significantly higher than the NCC values of all other images in the group. In group image #2, when the target image does not exist, a maximum normalized cross-correlation (NCC) value of 0.5638 is lower than the value obtained when the target image was included in the group image. Ultimately, a test was conducted to determine if the target image was present in the group image. The results indicated that the max NCC value was below 0.6, indicating no target image was identified. The designed recognition algorithm has proven successful in all cases.