Computer Science Colloquia
Wednesday, December 7, 2011
Advisor: Gabe Robins
Attending Faculty: Jason Lawrence (Chair), Wes Weimer, abhi shelat, Worthy Martin.
Olsson Hall, Room 228E, 3:00 PM
Ph.D. Qualifying Exam Presentation
Multispectral Camera Calibration for Image Based BRDF Acquisition
We present a carefully calibrated multispectral high dynamic range imaging system for image based acquisition of multispectral BRDFs. Our system allows us to take images under known illumination and extract the spectral reflectance at every pixel. Results are validated against objects of known spectral reflectance. Preliminary results are presented in using this multispectral camera for image based BRDF capture. My long-term goal with this project is to use the multispectral camera described in this quals document to capture a large database of densely sampled multispectral isotropic BRDFs. I then plan on using this database to verify analytical BRDF models, learn procedures for converting from RGB to multispectral BRDFS, and derive new analytical models. An analysis of a large database of multispectral images is also presented. We present results concerning finding optimal spectral sensors for multispectral imaging. Furthermore, discuss the compressibility of multispectral images using principal component analysis (PCA) and non-negative matrix factorization (NMF).