RUS  ENG
Full version
JOURNALS // Journal of Computational and Engineering Mathematics // Archive

J. Comp. Eng. Math., 2024 Volume 11, Issue 3, Pages 52–I (Mi jcem265)

Engineering Mathematics

Suppressing of image digital noise using a neural network based on U-Net

A. A. Kuznetsov

South Ural State University, Chelyabinsk, Russian Federation, kuznetsovaa@susu.ru

Abstract: A digital image is a computer representation of an optical image. The process of obtaining a digital image using digital cameras is always accompanied by noise. Removing noise from an image is an important stage in digital image processing, since noise at large values degrades the quality of the image and complicates subsequent analysis of the data on it. Noise in the image can occur due to environmental factors, ISO sensitivity, camera sensor and so on. The purpose of this research is to create a method for improving the visual quality of images by reducing the noise presented in them. This method, based on neural networks, will work with RAW images, converting them into RGB images. The resulting RGB image will be noise-free. The performance of the presented technique is evaluated in terms of noise reduction and image detail preservation. Experimental results demonstrate the effectiveness of the proposed denoising method in achieving significant noise reduction while maintaining image details.

Keywords: RAW Images, U-Net, image denoise.

UDC: 004.932

MSC: 68T07

Received: 25.12.2023

Language: English

DOI: 10.14529/jcem240305



© Steklov Math. Inst. of RAS, 2026