Approximation of the apparent bidirectional reflectance distribution function using asymmetric tensor splines,
efficient facial image relighting by 3rd-order polynomial evaluation,
enhancement of face recognition algorithms.
Collaborators
Arunava Banerjee, PhD
Ritwik Kumar, PhD
Baba C. Vemuri, PhD
Selected Publications
A. Barmpoutis, R. Kumar, B. C. Vemuri, and A. Banerjee. "Beyond the Lambertian Assumption: A generative model for Apparent BRDF fields of Faces using Anti-Symmetric Tensor Splines". CVPR, 2008
R. Kumar, A. Barmpoutis, A. Banerjee, and B. C. Vemuri. "Non-Lambertian Reflectance Modeling and Shape Recovery for Faces using Anti-Symmetric Tensor Splines"
IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, in press.
R. Kumar, A. Barmpoutis, B. C. Vemuri, and A. Banerjee. "Face relighting for recognition", Technical Report, CISE, University of Florida, 2008
This animation shows first the 9 input facial images along with their lighting directions.
Then, the 3rd-order tensor field is shown after having fitted a tensor spline to the image data. Notice the variability of the shapes of the local ABRD functions, which account for specularities and cast-shadows.
Finally, a demonstration of image relighting is shown, by evaluating the field of 3rd-order polynomials using various lignting directions.
