Title: Deep salient video deblurring framework using GAN

Authors: Arti Ranjan; M. Ravinder

Addresses: Department of Computer Science and Engineering, Indira Gandhi Delhi Technical University for Women (IGDTUW), New Delhi, India; Department of Computer Science and Engineering, Galgotias College of Engineering and Technology, Greater Noida, India ' Department of Computer Science and Engineering, Indira Gandhi Delhi Technical University for Women (IGDTUW), New Delhi, India

Abstract: Deblurring has been a major challenge in image restoration tasks. Video deblurring poses a new set of challenges as compared to image deblurring due to dimensionality as well as information connected in frames. GAN has been quite successful in the image deblurring tasks, taking inspiration from this we have proposed deep salient video deblurring (DSVD) framework using GAN. Saliency features help reduce the area of focus in the frames while guiding the GAN through the training phase. We have used video deblurring and GoPRO datasets to test our model and compared it to the various techniques we have discussed. In terms of peak signal-to-noise ratio (PSNR), our approach performs better than others. It particularly does well on heavily noised video frames. The results have been optimistic.

Keywords: deep salient video deblurring; DSVD; video deblurring; salient video deblurring; GAN.

DOI: 10.1504/IJCVR.2026.155556

International Journal of Computational Vision and Robotics, 2026 Vol.17 No.2, pp.159 - 177

Received: 12 Dec 2022
Accepted: 15 Nov 2023
Published online: 05 Aug 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article