An Inverse Tone Mapping Algorithm Based on Multi-scale Dual-branch Network

Abstract

Inverse tone mapping has drawn more attention recently because it can convert a large number of existing standard dynamic range (SDR) images into high dynamic range (HDR) images. In this paper, we proposed an inverse tone mapping algorithm based on a multi-scale dual-branch network which can restore the original information lost in under-/over-exposed areas. A multi-scale structure and masking mechanism are used to guide the reconstruction of image texture and structure. In order to enhance the robustness of the model for dealing with extremely exposed images, we apply a preprocessing method of exposure adjustment which improves the quality of the generated images. With quantitative and visual inspection experiments, we prove that the proposed algorithm has better performance than most state-of-the-art algorithms.