Image denoising using convolutional neural network

Abstract

Noise removal is one of the classic problems in low-level machine vision and image processing. After camera invention, image noise removal became necessary. Its purpose is to eliminate noise while preserving the edges and other precise and vital details as much as possible. Due to the outstanding results produced by deep learning in a variety of domains, this study introduces a novel model for image denoising that uses a mix of dilated and regular convolution to remove noise from degraded photos with a specified noise level. The proposed model is assessed on the BSD68 dataset, and the findings reveal that it outperforms models such as DnCNN and FFDNet at all noise levels.