Proposing an Enhanced Face Recognition Technique Using Multi- Wavelet Transform

Face recognition has been an active research for more than three decades. In general face recognition consists of feature extraction and classification. Principal Component Analysis (PCA) is one of the methods for features extraction under the appearance-based approach. However PCA came with large database, having some problems like load computation, and poor recognition accuracy. This research is an attempt to focus on large size of database problems with PCA technique. In order to make PCA work well, we suggest using MultiWavelet transform, which has two features. The first one is that it can decompose the image into frequency subband and the second feature is to reduce the resolution of the image. Moreover, in this research, the face recognition system can be classified into two phases. The first phase is training, which is of four steps. The first one is applying Multi-Wavelet transform to reduce the size of image. The second is PCA for feature extraction from images after reducing the size. The next is projection images, and finally is the library of projection images. The second phase is recognition, which is composed of three steps. The first step is similar to the training phase applied in the Multi-Wavelet transform. The second one is PCA for extraction the average from testing images, while the third is using the classification method via the Nearest Mean Classifier. Observations from the experiment results, based on Yale databases, show that the proposed method has better accuracy rate than the original PCA.