Evaluating tone mapping algorithms for rendering non-pictorial (scientific) high-dynamic-range images

Abstract

Nine algorithms were implemented to overcome the problem associated with rendering high-dynamic-range scientific imagery to low-dynamic-range display devices. The algorithms were evaluated using two paired-comparison psychophysical experiments judging preference and ”scientific usefulness”. The results showed that, on average, the Zone System algorithm performed best and the Local Color Correction method performed the worst. However, the performance of the algorithms depended on the type of data being visualized. The low correlation between the preference and scientific usefulness judgments ( R 2 = 0.31) indicated that observers used different criteria when judging the image preference versus scientific usefulness. The experiment was repeated using expert observers (radiologists) for an MR scan (Medical image). The results showed that the radiologists used similar criteria as the non-expert observers when judging the usefulness of the rendered images. A target detection experiment was conducted to measure the detectability of an embedded target in the Medical image. The result of the target detection experiment illustrated that the detectability of targets in the image is greatly influenced by the rendering algorithms due to the inherent difference in tone mapping among the algorithms.