Visual Models for Realistic Image Synthesis.

PhD thesis, Cornell University, 1998.

This thesis explores how psychophysically-based models of vision can be used to improve the fidelity and efficiency of image synthesis algorithms. In separate chapters computational models of visual adaptation, spatial vision, and spatial vision incorporating adaptation are developed and applied to problems in realistic image synthesis. The adaptation model addresses the problem of tone reproduction in image display and allows images of scenes illuminated at a wide range of absolute levels to be displayed within the limited ranges available on conventional display devices such as CRT's and printers. The spatial vision model is used to investigate the properties of visual masking, which can be used to predict the visibility of artifacts in synthetic images and used to choose surface textures for synthetic objects that hide these artifacts. Finally the model of adaptation and spatial vision returns to the issue of tone and color reproduction and provides a more comprehensive approach for displaying wide absolute range and high dynamic range color images on conventional displays. This work highlights the potential for useful symbiosis between the fields of perception psychology and computer graphics.