Systems evaluation for computer graphics rendering of the total appearance of paintings

One of the challenges when imaging paintings is recording total appearance, that is, the object\u27s color, surface microstructure (gloss), and surface macrostructure (topography). In this thesis, various systems were used to achieve this task, and a psychophysical paired comparison experiment was conducted to evaluate their performance. A pair of strobe lights arranged at 60° from the normal on either side of the painting captured color information where the strobes produced either directional or diffuse illumination geometry. By adding a third strobe, arranging them 120° apart annularly, and cross polarizing, diffuse color and surface normal maps were measured. A fourth strobe was added and the four lights were rearranged 90° apart annularly, capturing similar data. This system was augmented by two scanning linear light sources arranged perpendicularly, facilitating the measurement of spatially varying BRDF and specular maps. A laser scanner was used to capture surface macrostructure and was combined with the diffuse color maps from the four-light configuration. Finally, a dome illumination system was used with software developed by Conservation Heritage Imaging to produce color maps. In all, eight different configurations were achieved and used to image three small paintings with a range of appearance attributes. Twenty-five naive observers compared computer-graphic renderings to the actual painting and judged similarity in terms of total appearance, gloss/shininess, texture, and color. Although the rankings varied with painting, two general trends emerged. First, the four-light configuration with or without the independent laser scanning produced images visually equivalent to conventional strobe illumination. Second, diffuse illumination was always ranked lowest