Regional Weighted Generative Adversarial Network for LDR to HDR Image Conversion

Abstract

For the image conversion from low dynamic range images to high dynamic range images, we considered the possibility of using the Generative Adversarial Network (GAN) as one of the deep neural network methods. Deep learning requires a lot of data before building a module, but once creation is done, it is convenient to use it. In this paper, we proposed a weighting map for a local luminance based learning to reconstruct locally tone-mapped images.