An Efficient and Robust Combined Feature Extraction Technique for Face Recognition Systems

Abstract

Face recognition has long attracted a lot of interest from the research and market communities due to
its many possibilities across numerous sectors, but it has proven to be exceedingly difficult to deploy in real-time applications. Over the years, several face recognition algorithms and their variations have been created. In this paper, an integrating STIP and SURF for a robust feature extraction approach is proposed. This approach consists of four steps: In the first step, researchers are collecting the input images. In the next step, image preprocessing using a Gaussian filter is used. Then, image segmentation is applied using Region of Interest (ROI). The Spatial-Temporal Interest Point (STIP) is employed to extract the features related to facial behaviors from Facial Action Units (FAUs). The most effective approach for object recognition in image processing that applies feature descriptors is a Histogram of Oriented Gradients (HOG). In the last phase, use the feature selection process using SURF (Speeded-up Robust Features). This proposed approach achieved (0.25 ms) better performance than the traditional approach.