State of the Art in Perceptually Driven Radiosity

Abstract

Despite its popularity among researchers the radiosity method still suffers of some disadvantages over other global illumination methods. Besides the fact that the original method allows only for solving the global illumination of environments consisting of purely diffuse surfaces, the method is rather computationally demanding. In the search for possible speed-up techniques one of the possibilities is to take also the characteristic features of the human visual system. Being aware of how the human visual perception works, one may compute the radiosity solution to lower accuracy in terms of physically based error metrics, but being sure that the physically correct solution won’t bring any improvements in the image for the human observer. In the following report we briefly summarize achievements in the radiosity research in the past years and present the state of the art in perceptual approaches used in computer graphics nowadays. We will give an overview of known tone-mapping and perceptually-based image comparison techniques that can be used in the scope of the radiosity method to further speed up the computational process. In the second part of the report we concentrate on known radiosity methods that already use these perceptual approaches to predict different visible errors of the result of the radiosity computation. We will not speak about importance-driven radiosity solutions, as those methods are based on using geometric visibility rather than on using human perception-aware techniques.