Convolutional Neural Networks for Image Denoising in Infocommunication Systems
Олексій Іванович Шеремет, Kateryna S. Sheremet, Олександр Валентинович Садовой, Yuliya Sokhina · 2018
Image denoising is an urgent task that arises in image processing and transmission. Images are an integral part of content transmitted in infocommunication systems. Traditional statistical image filtering algorithms are not always effective for the random nature of the noise spectrum. Image denoising by convolutional neural networks is a modern and effective approach. The article is devoted to demonstrate the possibilities of using denoising convolutional neural networks to solve one of the most difficult tasks that developers face when performing the transfer of graphical information in infocommunication systems - denoising. Unlike traditional algorithms, denoising convolutional neural networks have architecture features that allow them to perform effectively image filtering with unknown noise level. It is suggested to use denoising convolutional neural networks to generate a correction signal in the infocommunication system, which transmits a noisy image. The article proposes to use denoising convolutional neural networks to generate a correction signal in the infocommunication system, which transmits a noisy image. Such pre-trained correction elements on a large and diverse image dataset can be easily fine-tuned for special filtering tasks.