Face video Super Resolution using Deep Convolutional Neural Network

In video surveillance, low resolution in face recognition is a major problem. Various Super Resolution (SR) approaches are introduced to perform the high resolution of face video recognition from low resolution videos. However, enhancing the resolution of face videos and reconstructing the high frequency data is a major problem in research area. Therefore, an effective face video Super Resolution method-based on Deep Convolutional Neural Network (Deep CNN) is introduced in this paper to achieve the face resolution effectively. Initially, the input video collected from the database is passed into the frame extraction stage, where the video frames are extracted and the face detection is carried out using the Viola Jones algorithm. Moreover, the detected image frame is processed by the Deep Convolutional Neural Network (Deep CNN) to enhance the image resolution. Deep CNN is highly effective in performing super resolution in face videos. However, the proposed Deep Convolutional Neural Network attains better performance using the metrics, like Second Derivative like Measure of Enhancement (SDME) as 0.9743 using video-1, and Feature SIMilarity index (FSIM) as 53.843 for video-4, respectively.