Statistical Models for Face Recognition System With Different Distance Measures

Face recognition is one of the challenging applications of image processing.Robust face recognition algorithm should posses the ability to recognize identitydespite many variations in pose, lighting and appearance. Principle ComponentAnalysis (PCA) method has a wide application in the field of image processing fordimension reduction of the data. But these algorithms have certain limitations likepoor discriminatory power and ability to handle large computational load. Thispaper proposes a face recognition techniques based on PCA with Gaborwavelets in the preprocessing stage and statistical modeling methods like LDAand ICA for feature extraction. The classification for the proposed system is doneusing various distance measure methods like Euclidean Distance(ED), CosineDistance (CD), Mahalanobis Distance (MHD) methods and the recognition ratewere compared for different distance measures. The proposed method has beensuccessfully tested on ORL face data base with 400 frontal imagescorresponding to 40 different subjects which are acquired under variableillumination and facial expressions. It is observed from the results that use ofPCA with Gabor filters and features extracted through ICA method gives arecognition rate of about 98% when classified using Mahalanobis distanceclassifier. This recognition rate stands better than the conventional PCA and PCA+ LDA methods employing other and classifier techniques.