Using the visual differences predictor to improve performance of progressive global illumination computation

A novel view-independent technique for progressive global illumination comp uting that uses prediction of visible differences to improve both efficienc y and effectiveness of physically sound lighting solutions has been develop ed. The technique is a mixture of stochastic (density estimation) and deter ministic (adaptive mesh refinement) algorithms used in a sequence and optim ized to reduce the differences between the intermediate and final images as perceived by the human observer in the course of lighting computation. The quantitative measurements of visibility were obtained using the model of h uman vision captured in the visible differences predictor (VDP) developed b y Daly [1993]. The VDP responses were used to support the selection of the best component algorithms from a pool of global illumination solutions, and to enhance the selected algorithms for even better progressive refinement of image quality. The VDP was also used to determine the optimal sequential order of component-algorithm execution, and to choose the points at which switchover between algorithms should take place. As the WP is computational ly expensive, it was applied exclusively at the design and tuning stage of the composite technique, and so perceptual considerations are embedded into the resulting solution, though no VDP calculations were performed during L ighting simulation. The proposed global illumination technique is also novel, providing interme diate image solutions of high quality at unprecedented speeds, even far com plex scenes. One advantage of the technique is that local estimates of glob al illumination are readily available at the early stages of computing, mak ing possible the development of a more robust adaptive mesh subdivision, wh ich is guided by local contrast information. Efficient object space filteri ng, also based on stochastically-derived estimates of the local illuminatio n error, is applied to substantially reduce the visible noise inherent in s tochastic solutions.