Abstract
A network structure (DRSN-GAN) is proposed for image motion deblurring that combines a deep residual shrinkage network (DRSN) with a generative adversarial network (GAN) to address the issues of poor noise immunity and low generalizability in deblurring algorithms based solely on GANs. First, an end-to-end approach is used to recover a clear image from a blurred image, without the need to estimate a blurring kernel. Next, a DRSN is used as the generator in a GAN to remove noise from the input image while learning residuals to improve robustness. The BN and ReLU layers in the DRSN were moved to the front of the convolution layer, making the network easier to train. Finally, deblurring performance was verified using the GoPro, Köhler, and Lai datasets. Experimental results showed that deblurred images were produced with more subjective visual effects and a higher objective evaluation, compared with algorithms such as MPRNet. Furthermore, image edge and texture restoration effects were improved along with image quality. Our model produced slightly higher PSNR and SSIM values than the latest MPRNet, as well as increased YOLO detection accuracy. The number of required parameters in the DRSN-GAN was also reduced by 21.89%.
Highlights
Image blurring can be caused by camera shaking or fast motion of a target object during exposure, which can be problematic in space exploration [1], facial recognition [2], video surveillance [3], and medical image recognition [4]
Experimental results showed the proposed deep residual shrinkage network (DRSN)-generative adversarial network (GAN) algorithm yielded the best restoration performance compared with similar deep network structures, such as DeblurGAN and DeblurGAN-v2
It achieved the best generalizability compared with the state-of-the-art algorithms, such as Deep Deblur, SRN, DeblurGAN, DeblurGAN-v2, the method of Suin et al, and MPRNet. erefore, the DRSN-GAN image restoration method proposed in this paper represents an improved image restoration algorithm
Summary
Image blurring can be caused by camera shaking or fast motion of a target object during exposure, which can be problematic in space exploration [1], facial recognition [2], video surveillance [3], and medical image recognition [4]. Motion deblurring techniques can be divided into traditional and deep learning methods. Conventional deblurring methods typically restore images by first estimating a motion blur kernel [5,6,7,8,9,10,11,12,13,14]. These approaches have limitations, such as complicated calculations, high noise levels, excessive blur kernel estimation requirements, heuristic parameter adjustments, and low generalizability, which often prevent their use in various scene types. This paper focuses on current popular deep learning methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
