Detail-revealing Deep Video Super-resolution

ICCV 2017  ·  Xin Tao, Hongyun Gao, Renjie Liao, Jue Wang, Jiaya Jia ·

Previous CNN-based video super-resolution approaches need to align multiple frames to the reference. In this paper, we show that proper frame alignment and motion compensation is crucial for achieving high quality results. We accordingly propose a `sub-pixel motion compensation' (SPMC) layer in a CNN framework. Analysis and experiments show the suitability of this layer in video SR. The final end-to-end, scalable CNN framework effectively incorporates the SPMC layer and fuses multiple frames to reveal image details. Our implementation can generate visually and quantitatively high-quality results, superior to current state-of-the-arts, without the need of parameter tuning.

PDF Abstract ICCV 2017 PDF ICCV 2017 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Video Super-Resolution MSU Video Upscalers: Quality Enhancement SPMC PSNR 26.99 # 39
SSIM 0.933 # 40
VMAF 51.96 # 12
Image Super-Resolution Set14 - 4x upscaling SPMC PSNR 27.57 # 70
SSIM 0.76 # 62
Video Super-Resolution Vid4 - 4x upscaling DRDVSR PSNR 25.88 # 14
SSIM 0.774 # 11

Methods


No methods listed for this paper. Add relevant methods here