GAUSSIAN PROBABILISTIC NON ADDITIVE ENTROPY BASED KERNEL ENTROPY COMPONENT ANALYSIS WITH SCALE INVARIANT FEATURE TRANSFORM FOR FACE RECOGNITION

The paper presents a methodology to perform face recognition using a Gaussian Probabilistic Non Additive Entropy based Kernel Entropy Component Analysis with Scale Invariant Feature Transform. The proposed technique adds invariance towards Pose, Illumination and Expression (PIE) changes in the face. The conventionally used Renyi entropy has been replaced with a Gaussian non-additive entropy measure for better representation of information content in the non-extensive systems containing some degree of regularity or correlation. Scale Invariant Feature Transform being a texture based local feature detector and descriptor that transforms an image into a large collection of local feature vectors each of which is reasonably invariant to image translation, scaling, rotation, partial occlusion and illumination further aids in adding robustness to the system.