RIS-GAN: Explore Residual and Illumination with Generative Adversarial Networks for Shadow Removal

摘要: Residual images and illumination estimation have been proved very helpful in image enhancement. In this paper, we propose a general novel framework RIS-GAN which explores residual with Generative Adversarial Networks for shadow removal. Combined the coarse shadow-removal image, estimated negative inverse maps can be used to generate indirect refine result fine shadow-free coarse-to-fine fashion. Three discriminators are designed distinguish whether predicted images, real or fake jointly compared corresponding ground-truth information. To our best knowledge, first one explore We evaluate proposed method on two benchmark datasets, i.e., SRD ISTD, extensive experiments demonstrate that achieves superior performance state-of-the-arts, although no particular shadow-aware components generators.