Video Super Resolution Techniques: A Survey
Neeboy Nogueira, Shawnon Guedes, Vaishnavi Mardolker, Amar Parab, Shailendra Aswale, Pratiksha R. Shetgaonkar · Zenodo (CERN European Organization for Nuclear Research) · 2021
Security plays a critical role in our lives. To protect homes, offices, valuable property we need to install security cameras. Most of the time criminals get away due to bad video quality. To overcome this problem, video super resolution technique is applied onto these feeds which can then store the enhanced video footage in turn increasing the chances of any criminal going unaccounted. High resolution video streaming devices can be replaced using the technology to upscale the already existing videos to a higher resolution for a better user experience. Video super resolution is achieved by recovering a high-resolution video from a low-resolution video. In this paper different existing techniques of video super resolution are surveyed and compared. It is found that deep learning technique Convolutional Neural Network (CNN) is a promising solution to achieve video super resolution. In addition, novel video enhanced technique is proposed to enhance the live security feed from a low-resolution security camera.