EnhanceNet: Single Image Super-Resolution Through Automated Texture Synthesis

Single image super-resolution is the task of inferring a high-resolution image from a single low-resolution input. Traditionally, the performance of algorithms for this task is measured using pixel-wise reconstruction measures such as peak signal-to-noise ratio (PSNR) which have been shown to correlate poorly with the human perception of image quality. As a result, algorithms minimizing these metrics tend to produce over-smoothed images that lack high-frequency textures and do not look natural despite yielding high PSNR values. We propose a novel application of automated texture synthesis in combination with a perceptual loss focusing on creating realistic textures rather than optimizing for a pixel-accurate reproduction of ground truth images during training. By using feed-forward fully convolutional neural networks in an adversarial training setting, we achieve a significant boost in image quality at high magnification ratios. Extensive experiments on a number of datasets show the effectiveness of our approach, yielding state-of-the-art results in both quantitative and qualitative benchmarks.

PDF Abstract ICCV 2017 PDF ICCV 2017 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Image Super-Resolution BSD100 - 4x upscaling ENet-E PSNR 27.50 # 25
SSIM 0.7326 # 28
Image Super-Resolution Set14 - 4x upscaling ENet-E PSNR 28.42 # 27
SSIM 0.7774 # 29
Image Super-Resolution Set5 - 4x upscaling ENet-E PSNR 31.74 # 34
SSIM 0.8869 # 35
Image Super-Resolution Urban100 - 4x upscaling ENet-E PSNR 25.66 # 26
SSIM 0.7703 # 26

Results from Other Papers


Task Dataset Model Metric Name Metric Value Rank Source Paper Compare
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling EnhanceNet FID 19.07 # 5
MS-SSIM 0.934 # 4
PSNR 29.42 # 3
SSIM 0.832 # 3
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling EnhanceNet FID 116.38 # 3
MS-SSIM 0.897 # 6
PSNR 23.64 # 4
SSIM 0.701 # 5

Methods


No methods listed for this paper. Add relevant methods here