Performance analysis of face recognition by combining multiscale techniques and homomorphic filter using fuzzy K Nearest Neighbour classifier

The face recognition problem is made difficult by the great variability in head rotation and tilt, lighting intensity and angle, facial expression, aging, partial occlusion (e.g. Wearing Hats, scarves, glasses etc.), etc. In this paper two multi scale techniques Discrete Cosine Transform and Discrete Wavelet Transform are used. Discrete Cosine Transform is applied by retaining various levels of DCT coefficients to face images prior to face recognition accuracy testing. Discrete Wavelet Transform is applied to face images and approximation coefficients at level 1 are used for face recognition. Homomorphic filter is used for illumination normalization. The aim is to find how the DWT and DCT coefficients when combined with the Homomorphic filter reduce the computational complexity. The complexity is reduced by either reducing the size of the image or by using the reduced feature set and how these techniques improve the face recognition rate. In this paper K Means clustering algorithm is used to cluster the pixels in face image. Binary threshold is applied in the clusters. The proposed work is to compare the performance of multiscale techniques DWT, DCT and by combining these multiscale techniques with Homomorphic filter using Fuzzy K Nearest Neighbour classifier by computing the face recognition accuracy rate. Face recognition accuracy is tested using the ORL face database