Multi-step face recognition for improving face detection and recognition rate

Abstract

Face detection at a distance is difficult owing to low-resolution images and blurring. For this reason, the traditional face recognition process has a low face detection rate and face recognition rate. Here, we propose a multi-step face recognition system, obtaining highresolution images using minimization and reliable regularization on the basis of bilateral filtering to deal with a variety of data and noise models. The main concept of our proposed system is to improve face recognition and detection rates using a multi-step face recognition process. The first step in our system is to use image restoration techniques on degraded and blurred images, eliminate noise, and apply superresolution algorithms after detection. The second step is to send the restored images from the first step to the face recognition process. The simulation results show that the face detection and recognition rates of our system are 74% and 54%, respectively. By contrast, these metrics for the single-step face recognition system are 23% and 8%, respectively. As a result, our proposed system produces better results than the single-step face recognition system. We theoretically justified and simulated our system with real-world outdoor video using a PTZ camera and a face recognition engine. Image identification before and after restoration is achieved using certain classification tools and methods. The experimental results demonstrate that our proposed system improved the recognition performance and the quality of the image. © 2016 Pushpa Publishing House, Allahabad, India.