Face Verification Using support Vector Machines with Histogram Intersection Kernal

Abstract

Face verification is an image categorization procedure. In this the face of the person is identified by using the given set of images. The precision of face verification system decreases when there is a variation in position of the testing image with the training images. The present thesis newlinediscusses various procedures to enhance the precision of face verification system under dissimilar orientations of testing and training images. One way of doing this is by adjusting the orientation of image applying bin calculation procedure. By being capable to modify the images, the testing and training images can be normalized so that they will have similar pose. After normalization training and testing images will have same pose. In this work we are utilizing support vector machines (SVM) newlinefor classification; the second procedure applies a Histogram intersection kernel planted in support vector discrimination function. The Histogram intersection kernel proved an enhanced the performance in face newlineverification. Utilizing these procedures, faces verification precision can be enhanced compared to other procedures that do not utilize pose adjustment or classic SVM kernels.