Seeing is Believing

By Anders Frick

How do we know if the image is real?

Alan Chalmers from the university of Bristol introduced the course "Seeing is Believing" in a quite dramatic way. He also gave a little background to the Sunday afternoon course in the West Hall A at the SIGGRAPH 2001 conference.

One of the main goals is to increase the realism and decrease the rendering time. Chalmers tells us that even if it is possible to allow very precise simulation of light-energy distribution in a scene, that doesn't ensure that the final image will have a high fidelity visual appearance.

Therefore another speaker, Ann McNamara from Trinity College in Dublin, Ireland, told the audience about how to determine the relationship between the physical world and how man understands that world. She also presented some important things to think about when you are going carry out experiment.

It is very important to think about the order effects. For example, if you always show an ugly picture before a nice picture, or always a dark image after a white one, there will surely be order effects that may affect the final test results. You must randomize in the "right way", she says.

McNamara teachs computer graphics and has been to the SIGGRAPH conference four times. She is a scientist but very interested in psychology and that was why she was speaking about psychophysical image quality metrics this year.

Further on, Scott Daly from Sharp Laboratories was talking. Daly is leader at the center of excellence for displayed appearance and also a key member of the technical staff at the digital video department at Sharp Labs. He talked about computational image quality metrics, which included everything from how the eye works to comparing different models such as the Visible Differences Predictor, VDP, and the Sarnoff model.  

Daly told the audience about the fact that human being is less sensitive to lines that have a 45-degree angle, and that knowledge has been used by the halftoning industry for long time. He also said that smooth tracking eye movements can reduce image velocity on the retina, and showed some examples on the big screen how people can perceive things, depending on many different factors.

Different people have simply different thresholds for what they can see and be aware of.

Seeing is believing and at last Holly Rushmeier from the IBM T.J. Watson Research Center talked about metrics and geometric simplifications. Image fidelity depends on reflectance, transmittance and emission from the image. She spoke about the limit for replacing geometry with just good texture. Texture detail can distract from simplified geometry and the knowledge how to find that limit between texture quality and geometry simplification is very valuable.