Photorealism in image synthesis usually needs a huge number of parameters and complex computations. It is successfull when an external observer is unable to see any difference between a photographic and a simulated picture. To achieve this goal, complex geometry, material surface and light-matter interaction simulations are needed.But there is an element that is often neglected in this simulation process : the visualisation environment. The simulated scene is viewed on a specific display, in a specific environment, by a human observer. Each one of these three components has specific caracteristics which could be exploited in image synthesis.Thus, we propose a perceptual image metric in order to perceptually compare two images by using a model which simulates the human vision process. We further integrate this metric into a progressive rendering algorithm, replacing the usual empiric parameters with perceptual ones. Computing times are then usually divided by two or three.
