神经网络在退化图像复原领域的进展综述

Advance of Neural Network in Degraded Image Restoration

退化图像复原是图像计算领域中的一个重要的难题。近年来以深度学习为代表的人工智能(AI)技术取得了快速的发展,越来越多的基于神经网络解决退化图像复原的研究工作出现。首先介绍了神经网络进行退化图像还原的主要技术并对图像复原的问题进行分类;然后利用神经网络解决退化图像复原问题中细分的多个主要问题,并对每个问题的当前研究现状与多种基于深度学习网络的解决方法的优势与局限性进行归纳分析,并给出与传统方法的对比。最后介绍了基于对抗神经网络的极限退化图像复原的新方法,并对未来前景进行展望。

Restoration of degraded image is an important and challenging issue in the field of image computing.In recent years,artificial intelligence(AI),especially deep learning,has achieved rapid progress.More and more methods based on neural networks have been proposed to solve this problem.This paper first introduces the main techniques based on neural networks to restore the degraded images and makes a classification of the problems.Then we focused on the key neural networks to resolve the problems of each category.By reviewing the development of various network-based methods in the field of deep learning,we analyzed the advantages and limitations between these methods.Furthermore,a comparison between these methods and the traditional ones was also made.Finally,we put forward a new solution on restoration of extremely degraded image using GANs,sketching out the future work on the restoration of degraded image.