Comparative Analysis of Rotation Invariant Pattern and Uniform Pattern in MMLBP Technique for Face Recognition

Abstract

Recognizing humans based on one or more physical or behavioral traits is referred as Biometrics. Comparing to the traditional methods to authenticate persons, biometric plays a vital role in the area of human recognition. In the field of biometric, face and palm print recognition seeks more attention for the researchers. The failure of recognition is minimum and also the implementation is easier than more other techniques. In this paper we concentrated on face recognition with Local Binary Pattern(LBP), it is simple and fast to recognize face than more other algorithms. Uniform LBP is used to extract features to recognize the face to authenticate the persons. The “non-uniform” patterns are clustered into one pattern due to this lot of information lost. In order to overcome the heavy data which loss in non-uniform patterns a modified multi-scale LBP histogram algorithm is proposed. Hence, the useful non-uniform information is utilized without any training step with entire information without any data loss. We also compare the mapping methods, rotational invariant pattern and uniform with rotational invariant patterns and .hence evaluate the performance of the mapping methods.