Reconstructing HDR Image from a Single Filtered LDR Image Base on a Deep HDR Merger Network

In this paper, a novel Deep HDR Merger network, which is called MergeNet, is proposed to reconstruct a HDR image from a single filtered LDR image. Filtered images are adopted as input since they contain more dynamic range than traditional ones. By learning the correlation between filtered LDR images and HDR images, the MergeNet successfully achieves HDR reconstruction of filtered images. We used five evaluation methods to make qualitative and quantitative comparisons to show that our method produced excellent results. Experimental results show that the proposed method performs favorably against state-of-the-art HDR image reconstruction methods.