Accurate appearance-based visualization of car paints

摘要

We propose a new method to generate images with a given color and texture, in order to visualize the appearance of car paints. Unlike current methods, the new method is based on visual comparisons of rendered paints with actual physical samples. Thus, we optimized the method to maximize the appearance match between rendered image and the corresponding car paint. In the new method it is possible to set accurate numerical values for not only color properties, but also for well-defined texture parameters. The new method is able to accurately render car paints under various light conditions, ranging from purely unidirectional, intense spot light to purely diffuse light. We show that the latter type of lighting conditions, which is often encountered in practical situations, is not well covered by existing rendering techniques that are based on BRDF and BTF measurements. Compared to existing methods for rendering, the proposed method is much faster regarding measurement and calculation, it has lower instrument costs and requires less data storage.