Confidence in Tone Mapping Applying a User-Driven Operator

Abstract

In photorealistic image synthesis, the natural appearance of a scene is predicted by simulating the illumination using radiometric values. Whenever the dynamic range of the simulated luminance values exceeds the capabilities of the display device, tone reproduction is necessary to reduce the contrast of the image. Although a signi cant number of tone mapping operators have been presented in the past, the reliability of the resulting low dynamic range images cannot be guaranteed. In the context of product design, decision-makers rely on a trustworthy colorimetric and photometric appearance. We believe that in a particular scenario a dedicated user-driven tone reproduction curve outperforms existing operators in terms of e xibility, performance, and quality. In this paper, we propose a method to manually generate a tone mapping operator. The user is guided to select a set of simulated input luminance values and to map them to appropriate display luminance output quantities. These key mappings are interpolated to a tone mapping curve. A module was developed for Qtpfsgui to de ne and apply the operator. We evaluated the resulting low dynamic range images in a study with thirteen participants. The probands were asked to directly compare a real with a virtual scene displayed on a low dynamic range device as well as to rate the results in comparison to popular tone mapping operators. In addition to de ning a dedicated curve for a speci c scenario, another application of our approach is to generate a standard observer tone reproduction curve by interpolating a set of user-driven functions.