EDGAN: motion deblurring algorithm based on enhanced generative adversarial networks
- PDF / 1,777,631 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 108 Downloads / 252 Views
EDGAN: motion deblurring algorithm based on enhanced generative adversarial networks Yong Zhang1,2,3 · Shao Yong Ma1 · Xi Zhang4 · Li Li5 · Wai Hung Ip6 · Kai Leung Yung6
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Removing motion blur has been an important issue in computer vision literature. Motion blur is caused by the relative motion between the camera and the photographed object. However, in recent years, some achievements have been made in the research of image deblurring by using deep learning algorithms. In this paper, an enhanced adversarial network model is proposed. The proposed model can use the weight of feature channel to generate sharp image and eliminate draughtboard artefacts. In addition, the mixed loss function enables the network to output high-quality image. The proposed approach is tested using GOPRO datasets and Lai datasets. In the GOPRO datasets, the peak signal-to-noise ratio of the proposed approach is up to 28.674, and DeblurGAN is 27.454. And the structural similarity measure can be achieved up to 0.969, and DeblurGAN is 0.939. Furthermore, the images were obtained from China’s Chang’e 3 Lander to test the new algorithm. Due to the elimination of the chessboard effect, the deblurred image has a better visual appearance. The proposed method achieved higher performance and efficiency in qualitative and quantitative aspects using the benchmark dataset experiments. The results also provided various insights into the design and development of the camera pointing system, which was mounted on the Lander for capturing images of the moon and rover for Chang’e space mission. Keywords Blurred image · Camera pointing system · Chang’e space mission · GANs · Resize convolution
* Li Li [email protected] Extended author information available on the last page of the article
13
Vol.:(0123456789)
Y. Zhang et al.
1 Introduction Motion blur is one of the most common types of image blurring. Shorter exposure time and fast-moving objects or camera shaking will cause motion blur in the final image, resulting in poor image perception, affecting image information transmission, and postprocessing [1, 2]. In the field of computer vision, motion blur could cause the reduction of accuracy and efficiency of image recognition and classification. Therefore, the restoration of motion blurred images is of great significance. Most of the early blurred image restoration approaches are based on the following blur model [3–6]:
IB = K ∗ IS + N
(1)
where IB , IS , K , and N are the blurred image, the latent sharp image, the blurred kernel, and noise, respectively, which represents a convolution operation. The process may be seen as a convolution operation between a sharp image and a blurred image kernel, when the blurred image is formed after the effect of random noise. An image restoration algorithm can be divided into blind restoration and non-blind restoration based on whether the blur kernel is known or not. The non-blind restoration algorithm restores the blurred image b
Data Loading...
