ACCV 2014, Tutorial
Color Transfer and Applications
Singapore
November 1, 2014

Course description



The numerical exercices will be done in Matlab, so please bring a computer with Matlab installed. The exercices and reference papers will be provided.


Color transfer is the problem of imposing the color palette of an image I1 on an image I2, without changing the spatial geometry of I2. It may be the case that the color distribution or the target characteristics are predefined, such as an equalized histogram, or an image without a colored- illuminant. Both problems are computationally challenging since they need to take into account the color and spatial domain information.


I1 I2 New image with
geometry of I1 and
colors of I2


Goal. This tutorial will present a state-of-the-art and unifying perspective of the color transfer problem between two or more images. We will also relate them to other important problems such as illuminant change or object recolorization. Our goal is to give a rigorous approach, that allows to understand the challenges of these problems, how the different assumptions made in the literature affect the final result, and how these problems are related together. Moreover, after the tutorial the audience should have an intuition of when and why the different algorithms in the literature can be applied, and the issues that still remain to be solved.

Length: 3h divided into two 1:30h units that include theory (45min each) and numerical guided exercices (45min each).

The numerical exercices will be done in Matlab, so please bring a computer with Matlab installed. The exercices and reference papers will be provided.




Authors - Biographical sketch of the proposers



Sira Ferradans (Vigo, 1982) received the B.Sc. degree in computer science, and the Ph.D. degree in image processing at the Universitat Pompeu Fabra, Spain in 2011. She was a postdoc associate at Paris-Dauphine for two years, and currently she works at Duke University. Her research interests include color image processing, HDR imaging, texture synthesis and mixing, and the application of optimal transport to imaging problems.


Marcelo Bertalmio (Montevideo, 1972) received the B.Sc. and M.Sc. degrees in electrical engineering from the Universidad de la Republica, Uruguay, and the Ph.D. degree in electrical and computer engineering from the University of Minnesota in 2001. Since 2006 he is an Associate Professor at Universitat Pompeu Fabra, Spain. His publications total more than 6,000 citations. He was awarded the 2012 SIAG/IS Prize of the SIAM organization of the USA for co-authoring the most relevant image processing work published in the period 2008-2012. Has received the Femlab Prize, the Siemens Best Paper Award, and the ICREA Academia Award, among other honors. He is an Associate Editor for SIAM-SIIMS and the secretary of SIAM's activity group on imaging. Has an ERC-Starting Grant for his project "Image processing for enhanced cinematography".





Target Audience



The topic is relevant for scientists, researchers and members of the industry community in very different areas such as computational photography, computer graphics, visual psychophysics, or computer vision. More specifically, the topics that we will develope can interest
The expected audience should have a basic knowledge of multivariate calculus and probability, although the course will be self-contained. Since we will be using code for the demos and exercices, the attendees should know some basic programming. We will assume no knowledge on the color transfer problem.




Topics and schedule



The numerical exercices will be done in Matlab, so please bring a computer with Matlab installed. The exercices and reference papers will be provided.

Covered material References and numerical exercices
First session(1:30h)

Histogram transfer, or global methods.
  • Presentation of the color transfer problem between two or more images.

  • 1D case: Histogram matching. Example of histogram equalization and relationship to Optimal Transport.

  • Full problem: 3D-distribution transfer and its computational challenges.

Second session(1:30h)

Taking into account the spatial domain and extension to several images


Global methods may not preserved the geometry of the images, to deal with this problem, several approaches have been proposed:
  • Statistical approach: The image is assumed to be a GMM in the spatial-color domain, then each of these Gaussians is transformed to match the Gaussians of the other image.

  • Variational approach: Change histogram with a spatial constraint.


  • Extension to several photographs and videos.


For more information : sira.ferradans.ramonde at duke.edu