GIT-GVU-91-13

Tone Reproduction for Realistic Computer Generated Images

Jack Tumblin
Holly E. Rushmeier

Radiosity and other global illumination methods for image synthesis calculate the 'real world' radiance values of a scene instead of the display radiance values that will represent them. This causes 'display range' problems that are often solved by ad-hoc means, giving little assurance that the evoked visual sensations (brightness, color, etc.) are truly equivalent. Workers in photography have studied such perception matching as 'tone reproduction', and devised correcting operators from both empirical and vision research data. Corrections are usually limited by the chemical/optical restrictions of film. These practical film methods were adopted by television systems and then by computer graphics, despite the ease of implementing better correction operators by computer. In this paper we advocate the use of better tone reproduction for computer graphics. We give a general framework for tone reproduction, where mathematical models of the display device and human observers define an explicit conversion from real-world to display device radiance. These are used to review tone reproduction operators used for film and television. A brief summary of some applicable vision research literature leads to a simple example of an improved operator. We apply the Stevens & Stevens models of brightness vs. luminance relations to our framework to create a new tone reproduction operator for black & white computer generated images. The new operator is shown to be a reasonable solution to the display range problem, and further extensions are suggested.

Computer imaging, radiance, tone reproduction

PDF

Postscript