Face Recognition Under Different Lighting Conditions

Face plays an important role in identification of any individual and it is the first and primary point of attention in social life. Face recognition is one of the important and popular visual recognition problems due to its challenging nature and its diverse set of applications. A face acknowledgment mechanism is one of the well-known Computer Vision Application for verification or identification of any individual either from a given picture or from any feature source. Face recognition finds application in surveillance, law enforcement, security and other areas. A face recognition is three-step process: preprocessing, feature extraction and classification. Before preprocessing, detection of facial part is carried out but now it is optional, in this, our aim is to select out the face region and reject out any other region so that further steps can be applied very effectively. Face detection is useful for compression, tracking of face, and estimating a pose of the face. Once a face portion is separated out then next step is to remove various disturbances present in that such as illumination variation, noise etc., using a preprocessing chain. Preprocessing chain consists of gamma correction, difference of Gaussian filtering, masking and contrast equalization. Then extricate the useful information using different feature extraction such as Local Binary Pattern (LBP), Local Ternary Pattern (LTP) and Dual Cross Pattern (DCP) from that in the form of feature vector. Local Binary Pattern compares the neighboring pixels with center pixel and generates the LBP code. In LTP, three-value code is generated depending upon the comparison between neighboring pixels and center pixel. From that, positive local binary pattern and negative local binary pattern are obtained. At last, system does recognition of face and returns the identity of that individual from the database. Accuracy measures, classification techniques and compression methods are involved in this phase. DCP encodes second order information in the vertical, horizontal and diagonal direction, by performing the encoding of sample points in the local surrounding region of every center pixel in an face image. Simulation is performed in Matlab 8.6 to evaluate the performance of face recognition framework using LBP,LTP and DCP, and ORL, AR, CMU-PIE face database is used for this purpose. The simulation performance of presented technique is evaluated in terms of face recognition rate. On the ORL face database, face recognition rate for LBP is 80.50%, for LTP is 87.50% and for DCP is 97.00% similarly on the AR database face recognition for LBP is 88.62% for LTP is 91.73% and for DCP is 95.74% On the CMU PIE database for LBP is 23% for LTP is 27.85% and for DCP is 73%