Hybrid WT Based-DCT Based Face Recognition

In this paper a robust face recognition algorithm based on both WT (wavelet transform) and DCT (discrete cosine transform) is presented. The proposed algorithm takes the advantage of data reduction property of both WT and DCT, hence a large reduction in the data size, while saving most of the information. Initially, 2D WT is used to compress the data at various levels, which also removes the high frequency noise from the input image. Then 1D DCT is applied to each column of the resulting image from the previous step for more data reduction. Finally, PCA (principle component analysis) is applied to extract principle components of the data in the training set. SVM (support vector machine) is then used to separate different classes (picture of different persons obtained at various conditions). Comparison of the proposed algorithm with previously reported WT based algorithm shows a significant improvement in the classification results.