Cyclic Learning-Based Lightweight Network for Inverse Tone Mapping

Recent studies on inverse tone mapping (iTM) have moved toward indirect mapping, which generates a stack of low dynamic range (LDR) images with multiple exposure values (multi-EV stack) and then merges them. In order to generate multi-EV stack(s), several large-scale networks more than 20 M parameters been proposed, but their high (HDR) reconstruction generation performance were not acceptable. Also, some previous methods using cycle consistency should even trained additional that are used for generation, results in large memory training. Thus, this paper proposes novel cyclic learning based reduce the burden detail, we eliminated only training, so proposed method enables efficient terms training-purpose memory. addition, presents lightweight iTM network dramatically reduces sizes existing networks. Actually, requires small parameter size 1/100 compared state-of-the-art (SOTA) method. The contributes practical use iTM. Therefore, reliably stack. Experimental show achieves quantitatively SOTA is qualitatively comparable conventional methods.