Image Quality Metrics
Advances in image synthesis allow very precise simulation of how light energy is distributed in a scene, but they do not ensure high perceptual fidelity. Contributing factors include: the limited dynamic range of displays, residual shortcomings of the rendering process, and the extent to which human vision encodes such departures from perfect physical realism. This course addressed techniques for comparing real and synthetic images, identifying important visual system characteristics, and significantly reducing rendering times.

Prerequisites
Basic understanding of realistic image synthesis and some knowledge of visual perception. Prior knowledge of image quality metrics not required.

Topics
Fidelity of images, general principles of human visual perception, human perception of lightness, psychophysical techniques, computational models of perception (spatial and orientation channels and visual masking), and computational metrics (visual difference predictors, the Sarnoff model, and animation quality metrics).

Organizers
Alan Chalmers
Ann McNamara
University of Bristol

Lecturers
Alan Chalmers
Ann McNamara
Tom Troscianko
University of Bristol

Scott Daly
Sharp Labs

Karol Myszkowski
University of Aizu