In this talk, I will introduce the problem of color transfer and how it is related to the well studied Optimal transport problem. Given any two images, the problem of color transfer is how to impose the color histogram of one image onto the other image while preserving its geometry and essential characteristics. Many solutions have been proposed in the literature, but the problem continues to be open for images with very different histograms. Discrete optimal transport, also known in the computer vision community as the earth moving problem, allows to formalize this problem. In this talk, we will go through the details of the discrete optimal transport, the challenges it poses for color transfer and, more generally, for other matching problems, and how and when it can be applied successfully. We will show some results where our proposal works correctly and the challenges that are still open in the color transfer problem.
Sira Ferradans (Vigo, 1982) received the B.Sc. degree in computer science at the Universitat Pompeu Fabra in Barcelona and her master in computer vision at the Universitat Autonoma de Barcelona. She obtained her Ph.D. in image processing in 2011 under the supervision of Vicent Caselles and Marcelo Bertalmio, for her research on Computational Photography, more specifically on the visualization of HDR images and the Demosaicking problem. She was a postdoc associate at Universite Paris-Dauphine for two years working with Gabriel Peyre on the application of Optimal Transport to imaging problems. She moved in 2014 to Duke University (USA) to work with Guillermo Sapiro, on color imaging problems. Her research interests include color image processing, HDR imaging, texture synthesis and mixing, and the application of optimal transport to imaging problems.