Taempeulriteu it based on the face region extraction method of the present invention, the image input step; With respect to the images using the non-conversion taempeulriteu not converted by using the matrix operation face candidate region extraction step; And targeting the extracted face region candidate, a face using taempeulriteu conversion to convert the input image by using the matrix operation region extraction step; It includes.; Here, non-converted taempeulriteu used in the face candidate area extraction step, the process speed compared to the relatively conversion taempeulriteu is faster by a low accuracy extraction method, for example, taempeulriteu configure the mean face image using a face image to learn .; Further it as the face candidate region extraction step, and performing matching to, in performing the taempeulriteu matching set remind difference weighted by the area of taempeulriteu using a non-conversion taempeulriteu difference weights remind set by taempeulriteu region, eyes from taempeulriteu area and assigns a higher weight to the mouth area.; In addition, the conversion taempeulriteu used in the face region extraction step is relatively as the processing speed is high, the accuracy extraction slow way compared to the non-conversion taempeulriteu is constituted by a wavelet transformation on the face image for learning, a conversion taempeulriteu the wavelet transform of the four partial frequency region are selected for both the vertical and horizontal low frequency region and high frequency region in the vertical axis direction that represents an edge in the horizontal direction in taempeulriteu.
展开▼