FUZZY FACE DETECTION

Abstract

Face Detection is an important part of face recognition. The face recognition is done after detecting a face in the still image or video. We proposedan approach for face detection in still color images which is capable of detecting face invariant to pose variations in the color image. The faces are detected in still image after segmentation of skin pixels in the input image. The segmentation of input image is done using Fuzzy C-means clustering technique. The clustering of modified color space Y’Cb’Cr’ is applied on Cr’ channel. The clusters those qualify for skin pixels are selected and rest are neglected. These cluster are selected if there minimum and maximum value lie between the thresholds. Detecting skin pixels gives the region of faces and other body parts and thus make easy to detect face in the image. These regions which are not faces are eliminated by fuzzy rules in the proposed approach and if some non faces are left over by these rules then a decision rule gives final output as it check the presence of eyes in the current region. The eye regions are detected in each left over region after applying fuzzy rules of elimination. After detecting faces the comparison is done with other standard techniques of face detection.