Face Detection In Color Image Sequences, And Follow-up Study

The problem of face detection derives from that of face recognition, which recently has emerged into an individual subject that attracts a wide range of researchers due to its broad application values in many fields such as video surveillance, content-based image retrieval, etc.With face detection being concentrated, in this thesis, some related issues like facial feature detection and face tracking are profoundly discussed. A general review of existing schemes on this particular topic is first studied, followed by two practical methods, differ-image-based one and skin-color-segmentation-based one. The former approach localizes faces roughly based on motion information in a single frame, while the later extracts faces accurately from still color images by use of the color and shape information. Then, the thesis extracts facial features in the interior of the face by using improved edge projection method. Face tracking based on improved color histogram is also presented for a series of standard image sequences. Finally, several further research aspects are pointed out based on the analysis of experimental results of above algorithms.