Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network

CVPR 2016 Wenzhe ShiJose CaballeroFerenc HuszárJohannes TotzAndrew P. AitkenRob BishopDaniel RueckertZehan Wang

Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution. In these methods, the low resolution (LR) input image is upscaled to the high resolution (HR) space using a single filter, commonly bicubic interpolation, before reconstruction... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Image Super-Resolution BSD100 - 4x upscaling ESPCN PSNR 27.02 # 26
Image Super-Resolution BSD100 - 4x upscaling ESPCN SSIM 0.7442 # 6
Image Super-Resolution BSD100 - 4x upscaling ESPCN MOS 2.01 # 4
Image Super-Resolution Set14 - 4x upscaling ESPCN PSNR 27.66 # 30
Image Super-Resolution Set14 - 4x upscaling ESPCN SSIM 0.8004 # 4
Image Super-Resolution Set14 - 4x upscaling ESPCN MOS 2.52 # 4
Image Super-Resolution Set5 - 4x upscaling ESPCN PSNR 30.76 # 27
Image Super-Resolution Set5 - 4x upscaling ESPCN SSIM 0.8784 # 27
Image Super-Resolution Set5 - 4x upscaling ESPCN MOS 2.89 # 4
Video Super-Resolution Ultra Video Group HD - 4x upscaling ESPCN Average PSNR 37.91 # 1
Video Super-Resolution Ultra Video Group HD - 4x upscaling bicubic Average PSNR 36.20 # 3
Video Super-Resolution Vid4 - 4x upscaling ESPCN PSNR 25.06 # 9
Video Super-Resolution Vid4 - 4x upscaling ESPCN SSIM 0.7394 # 5
Video Super-Resolution Vid4 - 4x upscaling ESPCN MOVIE 6.54 # 3
Video Super-Resolution Xiph HD - 4x upscaling bicubic Average PSNR 30.30 # 3
Video Super-Resolution Xiph HD - 4x upscaling ESPCN Average PSNR 31.67 # 1