图像视频质量增强是为了减轻或消除其在有损压缩过程中的质量损失，从而生成得到更接近无损的高质量图像视频。直方图均衡、灰度变换等传统的图像质量增强方法是直接对图像的像素值进行处理，低通滤波、高通滤波等是对经过傅里叶变换后的图像频谱成分进行处理。近年来随着深度学习方法的广泛运用，出现了大量基于卷积神经网络、生成对抗网络、长短时记忆网络等深度神经网络进行图像视频质量增强的方法。本文对近年来基于深度学习的图像视频质量增强方法进行全面的综述，分为图像和视频两类：图像质量增强包括基于卷积神经网络和生成对抗网络的方法，视频质量增强包括基于卷积神经网络，生成对抗网络和长短时记忆网络的方法。本文介绍了图像视频质量增强的经典工作，并总结了几种神经网络的实现过程，整理不同方法的数据库以及相应的实验结果对比。最后，本文分析了现有方法存在的不足，及其可能的发展方向。

Quality enhancement of compressed image/video is to reduce or eliminate its quality loss during lossy compression, and generates high-quality image/video which is closer to lossless ones. Traditional image and video quality enhancement methods, such as histogram equalization and gray level transformation, process the pixel values of the image directly; the low pass filtering and high pass filtering process the spectral components of the image after Fourier transform. With the widespread use of deep learning, a lot of neural networks, such as convolution neural networks, generative adversarial networks, and long-short term memory networks have been used to enhance the image and video quality. In this paper, the methods of image/video quality enhancement based on deep learning are reviewed: image-based methods can be divided into convolution neural network and generative adversarial network based methods; video-based methods includes convolution neural network, generative adversarial network, and long-short term memory based methods. This paper introduces numerous classical works of image and video quality enhancement, summarizes the implementation process of several neural networks, collates the databases of different methods, and compares their corresponding results. Finally, this paper also analyzes the problems existing in the existing methods and their possible development directions.