A Novel Face Matching Technique using Mean-Shift with Region Merging

We proposed a novel method for face matching from face image database. In our method we have taken set of face images so recognition decisions need to be based on comparisons of face image database. This paper presents an approach to region based face matching. The low level image segmentation method mean shift is used to divide the image into many small regions. As a popular segmentation scheme for color image, watershed has over segmentation as compared to mean-shift and also mean-shift preserves well the edge information of the object. The proposed method automatically merges the regions that are initially segmented by mean shift segmentation, effectively extracts the object contour and then, matches the obtained mask with test database image sets on the basis of color and texture. Extensive experiments are performed and the results show that the proposed scheme can reliably form the mask from the face image and effectively matches the mask with face image sets.