Color Based Face Recognition Using Template Based Approach

Abstract

Face recognition has gained  an  extensive attention from  research point of view. This area is still remained with many challenging problems in day to day  application. Many face ientification  algorithms have been implemented during past two decades. This paper provides a  template based technique for face recognition. Performance of the method considered in the paper is improved by considering  preprocessing steps such as image restoration, using Wiener filter to minimize noise and smoothing the degraded input images. Next face detection  step using skin color based segmentation for selecting only facial skin areas inorder to execute face recognition process. A template image database is used as traning set. Algorithm is tested over FERET database images. Experimental results showed that our algorithm is invariant to illumination change, change in pose, presence of salt pepper noise up to 0.05 density, inplane rotated images up to ±55 degrees. Accuracy of our algorithm is 100% in almost all input image cases.