Towards efficient motion-blurred public security video super-resolution based on back-projection networks

Kehua Guo,Haifu Guo,Sheng Ren,Jian Zhang,Xi Li
DOI: https://doi.org/10.1016/j.jnca.2020.102691
IF: 7.574
2020-09-01
Journal of Network and Computer Applications
Abstract:<p>In public security, video super-resolution has been an important technology for assisting users in effectively locating potential safety hazards. However, the video collected from video monitoring equipment is usually polluted by motion blur and noise, which results in low video quality after reconstruction. In this paper, we propose a super-resolution reconstruction model for motion-blurred video to reconstruct corresponding high-resolution video from motion-blurred low-resolution video. First, a single-frame image super-resolution model is designed based on the improved deep back-projection network, which primarily includes the design of a recursive method. Second, a multi-frame image super-resolution model is designed based on an optical motion algorithm for motion estimation and motion compensation of video frame sequences. Finally, according to the motion-blurred low-resolution video pollution condition, the de-blurring algorithm and multi-frame image super-resolution are combined to propose a super-resolution reconstruction model for motion-blurred video. Experiments show that the proposed motion-blurred video super-resolution model is superior to other contrasted algorithms in clear video and motion-blurred video reconstruction and has a better reconstruction effect and higher practical value.</p>
computer science, interdisciplinary applications, software engineering, hardware & architecture
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that in public - safety videos, factors such as motion blur and noise lead to low video quality, which affects the effectiveness and accuracy of video surveillance. Specifically, the paper proposes a motion - blur - video super - resolution reconstruction model based on Back - Projection Networks (BPNN), aiming to recover high - resolution clear videos from low - resolution videos with motion blur. This technology is of great significance for improving the video surveillance effect in the public - safety field. It can assist the police in observing suspects and video information more clearly, thus contributing to case analysis and public - safety management. The main contributions of the paper include: 1. Designing a recursive learning method based on iterative back - projection to improve the reconstruction quality of single - frame video images. 2. Combining the optical - flow algorithm (such as FlowNet2.0) to achieve motion estimation and motion compensation of video - frame sequences, and designing a multi - frame - image super - resolution model, which effectively utilizes the time - dimension information of videos and improves the quality of reconstructed video frames. 3. Combining DeblurGAN (Deblur Generative Adversarial Network) with multi - frame - image super - resolution, proposing a super - resolution reconstruction model BVDBPN suitable for motion - blur videos, achieving deblur reconstruction of low - resolution motion - blur videos and significantly improving the quality of reconstructed videos. Through these innovations, the model proposed in the paper performs better than other comparison algorithms in handling the super - resolution reconstruction tasks of motion - blur videos and has higher practical value.