A tone-mapping operator for road visibility experiments

Abstract

One may wish to use computer graphic images to carry out road visibility studies. Unfortunately, most display devices still have a limited luminance dynamic range, especially in driving simulators. In this paper, we propose a tone-mapping operator (TMO) to compress the luminance dynamic range while preserving the driver's performance for a visual task relevant for a driving situation. We address three display issues of some consequences for road image display: luminance dynamics, image quantization, and high minimum displayable luminance. Our TMO characterizes the effects of local adaptation with a bandpass decomposition of the image using a Laplacian pyramid, and processes the levels separately in order to mimic the human visual system. The contrast perception model uses the visibility level, a usual index in road visibility engineering applications. To assess our algorithm, a psychophysical experiment devoted to a target detection task was designed. Using a Landolt ring, the visual performances of 30 observers were measured: they stared first at a high-dynamic range image and then at the same image processed by a TMO and displayed on a low-dynamic range monitor, for comparison. The evaluation was completed with a visual appearance evaluation. Our operator gives good performances for three typical road situations (one in daylight and two at night), after comparison with four standard TMOs from the literature. The psychovisual assessment of our TMO is limited to these driving situations.