Details preservation inspired blind quality metric of tone mapping methods

Ke Gu,Guangtao Zhai,Min Liu,Xiaokang Yang,Wenjun Zhang
DOI: https://doi.org/10.1109/ISCAS.2014.6865186
2014-01-01
Abstract:High dynamic range (HDR) images are extremely meaningful, especially in the space and medical fields. For visualization of HDR images on standard low dynamic range (LDR) display devices, how to convert HDR to LDR images naturally becomes a valuable issue, which has aroused a variety of tone-mapping operators (TMOs). To compare different LDR images created by distinct TMOs, researchers have recently provided a subject-rated tone-mapped image database, and then developed a full-reference objective tone-mapped image quality index (TMQI) based on the measurement of multi-scale signal fidelity and statistical naturalness. Instead, the basic property of HDR images about details preservation is studied in this paper. With it, a natural inference is that higher-quality tone-mapped images are capable of displaying much more details. We therefore propose a blind quality metric by estimating the amount of details in images generated by darkening/brightening an original tone-mapped images. Experimental results on the above tone-mapped image database confirm that the proposed method, despite of no reference, is robust and statistically superior to the currently optimal full-reference TMQI algorithm, and remarkably outperforms state-of-the-art no-reference IQA metrics.
What problem does this paper attempt to address?