Just noticeable difference survey of computer generated imagery using normal maps

Michael Edward Hoerter, Purdue University

Abstract

Normal maps are widely used as a resource-efficient means of simulating detailed topology on 3D surfaces in the gaming, simulation, and film industries. However, as surface mesh density increases, it is unknown at what level of density these increases become no longer perceivable, and whether normal maps significantly affect this threshold. This study examined at what point participants were unable to discern differences between one level of mesh density and another using an adapted staircase model. Participants identified this threshold for five different organic character models. The averages of each of these thresholds were taken and compared against the results of a control group, which observed the same models without normal maps. The study found that the average threshold for discerning differences in level of detail occurred in the 3,000 to 14,000 polygon range for normal mapped models, and the 240,000 to 950,000 range for the control group. This analysis suggested that normal maps have a significant impact on the viewer's ability to discern differences in detail, and that developing graphics beyond the range of 3,000 to 14,000 polygons is unnecessary for organic character models when normal maps are used.

Degree

M.S.

Advisors

Whittinghill, Purdue University.

Subject Area

Computer science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS