静止階調画像を対象とした人物の顔領域の自動抽出

抄録

An image processing method to extract human facial regions whose positions, sizes and rotational angles are un-known from gray-scale still images is proposed in this paper. In this method, first a facial template is generated from several human facial images. Next, the facial template is magnified, rotated and overlapped with an original image, and the matching ratio between the template and the original image is calculated. Since there are a vast number of combinations for parameters of this transformation, Genetic Algorithm is employed to search quickly a parameter set which gives a good matching ratio. The proposed method can be applied to positioning, recognizing and counting of humans.