Tone reproduction for computer graphics using photographic principles

Since its inception, one of the major goals of computer graphics has been the quest for photorealism in the synthesis of computer generated imagery. In pursuit of this goal, the field has relied on a photographic metaphor whereby the transport of light is traced from a scene, through a camera and onto an image plane. In the final stage of the synthesis process, the tone reproduction stage, simulated illuminance values are converted to appropriate pixel values for eventual display. Most work in tone reproduction for computer graphics has focused on modeling of the human visual system. In this work, an alternate means of tone reproduction based on photography is described. The dissertation specifies a system that models individual components of a photographic system (including the camera, enlarger and photographic materials) and simulates the mechanisms by which photographic prints are created. The proposed system is validated through a suite of informal tests and formal experiments whereby simulation results are compared with actual photographic negatives and prints. Visual inspection of validation test results illustrates that the simulation can satisfactorily produce renderings that match closely with actual negatives and prints. Numerically, error differences in tone between simulation results and actual processed images range from 0–10%. The system enables tone reproduction to be applied to computer generated imagery using the same parameter space available to photographers. As such, the system can be used to experiment with different photographic parameters, immediately illustrating the results of a set of photographic choices. Application of the system in photographic education, rendering system validation, and digital cinema is also discussed.