Prediction of results from subjective evaluation of real-time-capable tone-mapping operators applied to limited high-dynamic-range images

Björn Annighöfer, Touraj Tajbakhsh, Rolf-Rainer Grigat

Abstract: The dynamic range of digital image sensors is growing faster than the reproducible dynamic range of standard display and print devices. In an assessment, we evaluated eight real-time-capable, high-dynamic-range (HDR) tone-mapping operators, using HDR images restricted to 12 bits per channel, and mapping to 8-bit low-dynamic-range (LDR) output data. In our survey, 51 test persons rated the subjective quality of LDR images of 15 scenes. For each scene, the operators were tested against linear tone mapping. From that survey, we derived three operators that performed better than linear mapping and evaluated their impact on exposure when using tone-mapping in an imaging chain. To mathematically predict the subjective quality of tone-mapped images, we tested five figures of merit against the empirical results of the survey. We propose a quality measure derived from the Naka-Rushton equation. We evaluate the proposed metric against four other image quality metrics, with our novel measure matching the subjective results best.