HFD-SRGAN: Super-Resolution Generative Adversarial Network with High-frequency discriminator

Abstract

The high-frequencies of images is very important both in keeping the edges and suppressing artifacts. To improve the performance of single image super-resolution (SISR) based on the SRGAN framework, we propose Super-Resolution Generative Adversarial Networks with high-frequency discriminator (HFD- SRGAN) by designing an additional discriminator for image's high-frequencies extracted by wavelets. Based on SRGAN, the image's high frequencies extracted by discrete wavelet transformations (DWT) were then introduced into GAN. Moreover, an additional discriminator for these high frequencies was built. Since the proposed model provides a direct and efficient way to locates and estimates the high frequencies of the reconstruction image, the visual effects of reconstructed the images can be improved with fewer computation costs. Experiments show that HFD-SRGAN has improved the visual effects of SRGAN when using the same generator network as SRGAN. The evaluation results show the performance of our method is equal to the state-of-the-art methods.