Additive Gaussian noise removal based on generative adversarial network model and semi-soft thresholding approach

• Published in: Multimedia tools and applications Vol. 82; no. 5; pp. 7757 - 7777
• Main Author: Khmag, Asem
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.02.2023; Springer Nature B.V
• Subjects: Adaptive learning; Algorithms; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Digital imaging; Feature maps; Generative adversarial networks; Image analysis; Image classification; Image restoration; Methods; Multimedia; Multimedia Information Systems; Neural networks; Noise reduction; Pixels; Random noise; Smoothing; Special Purpose and Application-Based Systems; Wavelet transforms; Generative adversarial networks; Second-generation wavelets; Image enhancement; Thresholding optimization; Image denoising
• DOI: 10.1007/s11042-022-13569-6
• ISSN: 1380-7501

In digital image analysis and processing field of study, noise reduction and suppression have been stated as a common query. However, it is mostly essential issue to demesne the fine edges and ridges and tiny texture while suppressing the noise in processing of the digital images. In order to avoid causing “Over-strangling” phenomenon, semi-soft thresholding model is exploited to classify the sharp edges of the contaminated images. In this study, a self-adjusting generative adversarial network GAN is utilized. This procedure is used to extract the fine edge of the noised digital images in order to improve the actual signal in the high frequency components where the main parts of the clean pixels may consider as noise pixels, and as a result delete the unwanted noise from the tested image that might cause over smoothing to the resulted images. In order to further denoise the contaminated digital image, adaptive learning GAN model throughout scoring machine is exploited. Therefore, it preserves the information of input image and feature maps, learns the correlation between global and local features, improves image restoration performance, and suppresses phenomena such as over-smoothing that tend to occur in wavelets-based denoising. The proposed method is an end-to-end network structure with CNN-based preprocessing methods. Experimental results demonstrate that, in comparison with state-of-the-art noise removal techniques, the proposed method has better visual quality, and the proposed method improves PSNR by 2.27 dB and 0.85 dB on average compared with state-of-the-art- denoising methods. In addition, the proposed method could shorten the processing time noticeably.