Image Super-Resolution Based on Dense Convolutional Network
Recently, the performance of single image super-resolution (SISR) methods have been significantly improved with the development of the convolutional neural networks (CNN). In this paper, we propose a very deep dense convolutional network (SRDCN) for image superresolution. Due to the dense connection, the feature maps of each preceding layer are connected and used as inputs of all subsequent layers, thus utilizing both low-level and high-level features. In addition, residual learning and dense skip connection are adopted to ease the difficulties of training very deep convolutional networks by alleviating the vanishing-gradient problem. Experimental results on four benchmark datasets demonstrate that our proposed method achieves comparable performance with other state-of-the-art methods.
Single image super-resolution Dense convolutional network Residual learning
Jie Li Yue Zhou
Image Processing and Pattern Recognition,Shanghai Jiao Tong University,Shanghai,China
国际会议
广州
英文
134-145
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）