Deep convolution network for surveillance records super-resolution

• Published in: Multimedia tools and applications 2019-09, Vol.78 (17), p.23815-23829
• Main Authors: Shamsolmoali, Pourya, Zareapoor, Masoumeh, Jain, Deepak Kumar, Jain, Vinay Kumar, Yang, Jie
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Computer Communication Networks; Computer Science; Convolution; Data Structures and Information Theory; Image resolution; Machine learning; Model testing; Multimedia Information Systems; Neural networks; Special Purpose and Application-Based Systems; Surveillance
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-018-5915-7

The aim of image super resolution (SR) is to recover low resolution (LR) input image or video to a visually desirable high-resolution (HR) one. The task of identifying an object in surveillance records is interesting, yet challenging due to the low resolution of the video. This paper, proposed a deep learning method for resolution recovery, the low-resolution objects and points in the surveillance records are up-sampled using a deep Convolutional Neural Network (CNN) to avoid problems of image boundary the data padded with zeros. The network is trained and tested on two surveillance datasets. Dissimilar to the outdated methods which operate components individually, our model performs combined optimization for all the layers. The proposed CNN model has a lightweight structure and minimal data pre-processing and computation cost. Testing our model and comparing with advanced techniques, we observed promising results. The code is accessible at https://github.com/Mzareapoor/Super-resolution