Face Recognition Techniques Based on 2D Local Binary Pattern, Histogram of Oriented Gradient and Multiclass Support Vector Machines for Secure Document Authentication
Melkye Wereta Tsigie, Rasika Thakare, Rahul Joshi · 2018
Face recognition, one of the biometric computer vision research area, is a pattern recognition problem which have been done a dozen times since 1960s and it is still a revolutionary area of research interest for many researchers. Although face recognition is the earliest pattern recognition problem yet its accuracy is not as high as other biometric recognition problems like finger print recognition. Different imaging conditions made it challenging like occlusion of faces by hands or eye-glasses, illumination changes, variation in pose and different facial expressions. In this paper we proposed a robust face recognition technique by using local binary pattern and histogram of oriented gradient feature extractor and descriptors. The work has been conducted by carefully acquired and pre-processed 1300 face images, out of it 1040 images were used for training and the rest 260 images for testing purposes. As LBP operators are not good for extracting edge features of a face image we used HOG to extract edge features and LBP for extracting texture features of a face and finally the extracted features has been trained and classified by using multiclass support vector machines and it has shown good accuracy rate of recognition.