Post-filtering for Color Appearance in Synthetic Images Tone Reproduction

In the Photorealistic Image Synthesis process the spectral content of the synthetic scene is carefully reproduced, and the final output contains the exact spectral intensity light field of the perceived scene. This is the first important step toward the goal of producing a synthetic image that is indistinguishable from the actual one, but the real scene and its synthetic reproduction should be studied under the same conditions, in order to make a correct comparison and evaluate the degree of photorealism. To simplify this goal, a synthetic observer could be employed to compensate differences in the viewing conditions, since a real observer can not enter into a synthetic world. Various solutions have been proposed to this end. Most of them are based more on perceptive measures of the Human Visual System (HVS) under controlled conditions rather than on the HVS behavior under real conditions, e.g. observing a common image and not a controlled black and white striped pattern. Another problem in synthetic image generation is the visualization phase, or tone reproduction, whose purpose is to display the final result of the simulation model on a monitor screen or on a printed paper. The tone reproduction problem consists of finding the best solution to compress the extended dynamic range of the computed light field into the limited range of the displayable colors. We would like to propose a working hypothesis to solve the appearance and the tone reproduction problems in the synthetic image generation, integrating the Retinex model into the photorealistic image synthesis context, including in this way a model of the human visual system in the synthesis process.