Denoising the Endoscopy Images of the Gastrointestinal Tract Using Complex-Valued CNN

To delineate intestine atrophy accurately is a tough job because of the heterogeneous and convoluted structure of small intestine. Image quality is a salient factor to diagnose various diseases in the medical field. The images captured through machines or medical imaging devices are prone to some kind of noise. Endoscopy images are of inferior quality because of lighting problems inside the gastrointestinal (GI) tract. The noise type changes with the environment and the camera is used for capturing images. The aim of this paper is to determine whether image-denoising methods are effective for classification. This study proposes an efficient complex-valued CNN (CDNet) method for denoising the images. The proposed model found to be superior to any other state-of-the-art method for image denoising on real datasets. The denoising performance is computed through metrics of PSNR and SSIM. The results show PSNR 45.58 and SSIM 0.99 thus demonstrating the superiority of the proposed method on real datasets.