Verification of the Accuracy of the Luminance-Based Metrics of Contour, Shape, and Detail Distinctness of 3D Object in Simulated Daylit Scene by Numerical Comparison with Photographed HDR Images

Abstract

ABSTRACTThis study verifies the accuracy of previously developed luminance-based metrics of light modeling (that is, the distinctness of contour, shape, and details of daylit 3D objects) through comparison of numerical values of the metrics obtained from photographed and simulated high dynamic range (HDR) images. The analysis of the luminance data of eight photographed and eight simulated HDR images of Venetian masks showed that the mean relative error of all tested metrics was 14.78%. The minimum average relative error was 7.91%, and the maximum error (found for only one metric) was 27.75%. The glossy objects had higher error rates than matte objects tested within the experiment. Additionally, the variation among luminance-based metrics obtained from simulated and photographed HDR images was compared with variation among illuminance-based metrics. It became evident that luminance-based metrics (consequently color and specularity dependent) showed considerable variation among examined masks, affirming the impact of color and specularity on visibility of contour, shape, and detail. Though the tested luminance-based metrics may need further investigation in experimental conditions that differ from those used in this study, it is clear that they could be applied both with photographed and simulated luminance maps with precision of about 15%. Experimental results confirm the robustness of the metrics and encourage use of the luminance-based light modeling metrics in computer lighting simulations.