Class Discussion Summary (Feb 23)

Put the citations to papers discussed here:

Acquiring the Reflectance Field of a Human Face

http://www.pauldebevec.com/Research/LS/

Scribes:

Alex Salman - Group 5

Christian Lei

Kyle Oda

Rashmi Chennagiri

Melanie Wong

Write the Class Discussion Summary Here:

Q0: Based on the written discussions, there seems to be some confusion about how the "relighting" works. How is the new lighting condition determined? (not the change in viewpoint, just the change in lighting)

In this paper, they were able to extrapolate the reflectance from the human faces using a function to generate an image of the face in a novel form of illumination. For each color channel, they use the function and the illumination data gathered. This function involved taking the sampling from the reflectance function output and given an x,y coordinate pair and illumination direction phi and theta. This then corresponds to a linear summation that takes these inputs and outputs the relit image. 

Q1: This paper measures 2000 lighting directions with a giant device. Do you think you could measure fewer lighting directions? If so, what would you do for in-between lighting directions? Are there any properties of the measured data that make you think it would/wouldnt work to capture fewer lights?

We don’t think that the properties of the measured data would help capture better new viewpoints using fewer lights. And if it happened there will be some artifacts. There was also the idea that with fewer lights, we could calculate the missing angles. We also thought that some angles could not be interpreted without more exposure. We also thought that with specular reflections, it could create a washed out effect.