Laplacian Generative Adversarial Networks for Multi-Scale Super-Resolution

Single-image super-resolution, aims to reconstruct images from low-resolution to high-resolution and super-high-resolution, so as to improve image clarity and visual effects. Based on the principles of generating adversarial networks and image pyramids, a LapSRGAN is proposed to compromise the merits of Laplacian Pyramid Generative Adversarial Network(LapGAN) and Laplacian Pyramid Super-Resolution Network(LapSRN). LapSRGAN is an end-to-end image reconstruction network, which can achieve double and quadruple high-quality, high-resolution reconstruction of the original image. The proposed LapSRGAN is trained with multiscale discriminators and perceptual loss which calculated on feature maps of image. Extensive quantitative and qualitative evaluations on bench mark datasets confirm the effectiveness of the proposed algorithm.