High Dynamic Range Image Tone Mapping: Literature review and performance benchmark

Compared to conventional digital imaging, high dynamic range (HDR) image tone mapping technology can faithfully reproduce real-world scenes on common displays to provide a better viewing experience. The research on HDR image tone mapping has evolved over three decades and has been widely employed in many applications like photography, movie, and animation rendering. Nowadays, HDR image tone mapping is still an active area, especially with the successful use of deep neural networks in this field. This paper presents a literature review and constructs benchmarks on the existing tone mapping methods to pave the way for further developments. While there are previous reviews summarizing tone mapping algorithms, they are getting outdated as novel methods appear constantly. Therefore, we propose an updated literature review that includes state-of-the-art image tone mapping works to fill the gap. The reviewed methods are discussed and categorized based on the processed image channels and employed models. Furthermore, we summarize available HDR image datasets and objective metrics used to evaluate the quality of tone-mapped images. Two benchmarks are constructed to assess the reviewed tone mapping algorithms based on two large-scale datasets and seventeen quality metrics. Visual test is also conducted to evaluate the reviewed methods subjectively. This review concludes with a discussion of challenges that remain to be addressed and possible future research directions in HDR image tone mapping.