Object Recognition Based on Optimized Gabor Features and SVM

This paper proposes an optimized Gabor features and SVM based framework for object recognition. When discriminative features are extracted at optimized locations using tuned Gabor wavelets, classifications are done via SVM. Compared to conventional Gabor feature based object recognition system, the system developed in this paper is both robust and efficient. The proposed framework has been successfully applied to two object recognition applications, i.e. face/non-face classification and face recognition. Experimental results clearly show advantages of the proposed method over other approaches.