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

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. This means that the super-resolution (SR) operation is performed in HR space. We demonstrate that this is sub-optimal and adds computational complexity. In this paper, we present the first convolutional neural network (CNN) capable of real-time SR of 1080p videos on a single K2 GPU. To achieve this, we propose a novel CNN architecture where the feature maps are extracted in the LR space. In addition, we introduce an efficient sub-pixel convolution layer which learns an array of upscaling filters to upscale the final LR feature maps into the HR output. By doing so, we effectively replace the handcrafted bicubic filter in the SR pipeline with more complex upscaling filters specifically trained for each feature map, whilst also reducing the computational complexity of the overall SR operation. We evaluate the proposed approach using images and videos from publicly available datasets and show that it performs significantly better (+0.15dB on Images and +0.39dB on Videos) and is an order of magnitude faster than previous CNN-based methods.

PDF Abstract CVPR 2016 PDF CVPR 2016 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 4x upscaling ESPCN PSNR 27.02 # 47
SSIM 0.7442 # 15
MOS 2.01 # 4
Video Super-Resolution MSU Video Super Resolution Benchmark: Detail Restoration ESPCN Subjective score 2.099 # 31
ERQAv1.0 0.521 # 31
QRCRv1.0 0 # 21
SSIM 0.811 # 23
PSNR 26.714 # 18
FPS 3.333 # 2
1 - LPIPS 0.765 # 28
Video Super-Resolution MSU Video Upscalers: Quality Enhancement ESPCN PSNR 26.25 # 45
SSIM 0.926 # 36
VMAF 47.19 # 14
Image Super-Resolution Set14 - 4x upscaling ESPCN PSNR 27.66 # 68
SSIM 0.8004 # 6
MOS 2.52 # 4
Video Super-Resolution Ultra Video Group HD - 4x upscaling bicubic Average PSNR 36.20 # 5
Video Super-Resolution Ultra Video Group HD - 4x upscaling ESPCN Average PSNR 37.91 # 3
Video Super-Resolution Vid4 - 4x upscaling ESPCN PSNR 25.06 # 18
SSIM 0.7394 # 15
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

Methods