基于深度卷积网络的图像超分辨率重建

Learning a deep convolutional network for image super-resolution reconstruction

提出一种处理单幅图像超分辨率重建的深度学习方法,该方法通常把低分辨率图像作为输入,高分辨率图像作为输出,同时直接学习低/高分辨率图像之间的端到端映射关系,从而根据映射关系建立深度卷积神经网络(deep convolutional neural network,CNN)。另外,传统的基于稀疏编码的超分辨率重建方法同样可以认为是一种深度卷积网络,但是通常需要分别执行每一步操作,而文中方法则实现共同优化所有的卷积层。实验结果显示,深度卷积神经网络虽然结构简单,但是能够得到高质量的重建结果,同时在实际应用中实现高速运行。

A deep learning method for single-image super-resolution reconstruction is proposed,which usually takes a low-resolution image as an input and a high-resolution image as an output,and directly learns the end-to-end mapping between low-resolution and high-resolution images.A deep convolutional neural network(CNN)is established based on the mapping relationship.In addition,the traditional sparse-coding-based super-resolution reconstruction method can also be considered as a deep convolutional network,but usually needs to perform each step separately,and the method achieves the common optimization of all convolutional layers.The experimental results show that our deep convolutional neural network,although simple in structure,can obtain high-quality reconstruction results and achieve high-speed operation in practical applications.