Image Super-Resolution Using Deep Convolutional Networks

31 Dec 2014  ·  Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang ·

We propose a deep learning method for single image super-resolution (SR). Our method directly learns an end-to-end mapping between the low/high-resolution images. The mapping is represented as a deep convolutional neural network (CNN) that takes the low-resolution image as the input and outputs the high-resolution one. We further show that traditional sparse-coding-based SR methods can also be viewed as a deep convolutional network. But unlike traditional methods that handle each component separately, our method jointly optimizes all layers. Our deep CNN has a lightweight structure, yet demonstrates state-of-the-art restoration quality, and achieves fast speed for practical on-line usage. We explore different network structures and parameter settings to achieve trade-offs between performance and speed. Moreover, we extend our network to cope with three color channels simultaneously, and show better overall reconstruction quality.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 4x upscaling SRCNN PSNR 26.9 # 48
SSIM 0.7101 # 44
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling SRCNN FID 31.84 # 7
MS-SSIM 0.924 # 7
PSNR 27.40 # 6
SSIM 0.801 # 7
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling SRCNN FID 147.21 # 7
MS-SSIM 0.900 # 5