Accurate Image Super-Resolution Using Very Deep Convolutional Networks

CVPR 2016  ·  Jiwon Kim, Jung Kwon Lee, Kyoung Mu Lee ·

We present a highly accurate single-image super-resolution (SR) method. Our method uses a very deep convolutional network inspired by VGG-net used for ImageNet classification \cite{simonyan2015very}. We find increasing our network depth shows a significant improvement in accuracy. Our final model uses 20 weight layers. By cascading small filters many times in a deep network structure, contextual information over large image regions is exploited in an efficient way. With very deep networks, however, convergence speed becomes a critical issue during training. We propose a simple yet effective training procedure. We learn residuals only and use extremely high learning rates ($10^4$ times higher than SRCNN \cite{dong2015image}) enabled by adjustable gradient clipping. Our proposed method performs better than existing methods in accuracy and visual improvements in our results are easily noticeable.

PDF Abstract CVPR 2016 PDF CVPR 2016 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution IXI VDSR SSIM for 2x T2w 0.9836 # 7
PSNR 2x T2w 38.65 # 7
SSIM 4x T2w 0.9240 # 6
PSNR 4x T2w 30.79 # 6
Video Super-Resolution MSU Video Upscalers: Quality Enhancement VDSR PSNR 25.89 # 47
SSIM 0.917 # 35
VMAF 36.46 # 15
Image Super-Resolution Set14 - 2x upscaling VDSR [[Kim et al.2016a]] PSNR 33.03 # 20
Image Super-Resolution Set5 - 2x upscaling VDSR [[Kim et al.2016a]] PSNR 37.53 # 23
Image Super-Resolution Urban100 - 2x upscaling VDSR [[Kim et al.2016a]] PSNR 30.76 # 17

Results from Other Papers

Task Dataset Model Metric Name Metric Value Rank Source Paper Compare
Image Super-Resolution VggFace2 - 8x upscaling VDSR PSNR 22.50 # 4
Image Super-Resolution WebFace - 8x upscaling VDSR PSNR 23.65 # 4