FP-JamesOshea
From CS294-10 Visualization Fa07
Contents |
[edit] Proposal
What is the angle of the assumed light source for a scene?
[edit] Group Members
- James O'Shea
[edit] Description
Lighting plays an important role in revealing the 3D shape of an object. The proper lighting can effectively convey information about the surface structure and detail of an object, and ultimately have a significant effect on the viewer's perception of the scene. In computer graphics, lighting design is an integral part of the rendering process in order to produce effective shading. There have been many papers describing automatic methods for positioning lights in a scene in order to maximize the amount of shape information available to the viewer. These methods tend to use different metrics for assessing effectiveness, yet they rarely account for any assumptions the viewer may have about the scene. One particular assumption is the location of the light position. Where the viewer assumes the light source to be can significantly affect the shape of the surface they perceive. There is a large literature concerning the location of the assumed light source (generally referred to as the light-from-above prior), yet these models tend to be relatively simplistic, limiting the position of the assumed light source in a general 'above' or 'left' direction, or at most locating a degree bearing offset from vertical. It is not known, however, if the light is assumed to be at a particular angle from the viewing position. This angle is equivalent to the extent to which the light is grazing the surface of the object.
In this project, I will design a psychophysics study to shape perception and examine whether there is an assumed light position at a particular angle from the view position. The experiment will be a variation of the gauge figure studies made popular by Koenderink (Koenderink 1992). The basic task will be to position a gauge figure such that it appears normal to the surface of the object. By sampling multiple sites across a variety of shapes, we will able to assess how people perceive local surface orientations. In order to examine the assumed light position, we will vary the angle of the light incident on the surface across the trials and then analyze the resulting data to see if performance was influenced by the light position. In particular, we would like to see if precision and accuracy increase at a particular light position. If the viewer assumes the light to be in a particular position, then it stands to reason that their performance will be most reliable and accurate when the actual light position is consistent with this prior.
