Departmental Bulletin Paper 遺伝的アルゴリズムを用いた顔画像検出処理の改良

林, 晃平  ,  丸山, 晃生  ,  中村, 孝史  ,  荒川, 正和  ,  河原林, 友美

This paper describes a face image detection method which is executed by image processing immediately. Face image detection is one of the most mature studies in computer vision and has been gradually advanced practical use. But it has been still left room for studies on detection accuracy and processing speed. In our study, we applied an improved genetic algorithm into template pattern matching which is one of pattern recognitions, and we detected a face area in static images rapidly. In comparison with the template face image and a subject image, this algorithm increase comparison pixels as the generation in genetic algorithm increase. Furthermore, we adopt YCrCb color system for our face detection system in order to detect face images with relatively accuracy. As a result, we could mplement our face detection algorithm in motion movies. Finally, this paper reports the detection rate and the detection time for our face detection system.

Number of accesses :  

Other information