Spatial Domain Face Recognition System Using Convolution of PDV and LBP

Face recognition has become the new captivating field for scientists and researchers the world over. This paper, proposes an algorithm based on the convolution of the Pixel Difference Vector (PDV) and Local Binary Pattern (LBP) features. The features from the two techniques are convolved to generate a square matrix, which is then reshaped into a column vector. The column vectors of all the images that are present in the database are compared against the column vectors of the test image, making use of Euclidean Distance (ED). Following this, the location of the image in the database is obtained to detect the person and minimum distance between the specific image and the test image. The location is tracked so as to ensure precision. The results are used for matching, calculation of FAR, FRR and TSR. The model that has been proposed has been evaluated on the ORL database, JAFFE database, Indian Females database etc. The experimental results indicate that the systems proposed outperform the existing ones based on individual feature techniques and models employing multiple feature types.