Composite Feature Extraction based on Gabor and Zernike Moments for Face Recognition

This paper presents the experimental evaluation of Gabor filter and Zernike moments for extracting the face features. The dimensionality of the input image is reduced for the overloading process of Gabor filters. 40 sub-images were obtained from the original images by using the Gabor filters in 5 scales and 8 orientations. From each sub-image, four Zernike features were extracted. Thus, the total numbers of features are 160. The kNearest Neighbor (k-NN) classifier is used for the matching purpose. The experiments were performed in the ORL and NC-Face database of Facial Expressions. The recognition rate in the ORL database is 98.5% and the rate in the NC-Face database of Facial Expression is 89.23%. In the proposed system, the performance was found to be satisfactory as compared to the existing system.