A short resume of my carrer

I received the Bachelor's degree in Telecommunications Engineering from the Universitat Politècnica de Catalunya (UPC) in September 1996. Short afterwards I initated my PhD in the Image Processing Group of the UPC and obtained the degree in June 2002. In February 2003 I joined the Image Processing Group of the Department of Information and Communication Technologies at the Universitat Pompeu Fabra (UPF), where I held a "Ramon i Cajal" grant during the period 2003-2008 and a lecturer ternure-track position during 2008-2010. Since May 2010 I'm associate professor in the Universitat de Barcelona. I belong to the Computer Vision and Data Science Group.

Research interests

The research conducted so far has been concentrated on many different topics which are described next:

  • The first line of research was performed in the UPC, during the doctorate program. Graph structures provide a tool to represent the image at the region level and thus allow to perform a first level of abstraction of the image reducing thus at the same time the number of elements that have to be processed with respect the pixel level.

  • The second main research topic, performed at the UPF, has been focused on non-linear multiresolution and multigrid numerical optimization. Currently most approaches to minimize a continuous functional is based on computing the Euler-Lagrange equations of the system to be solved and on discretizing the resulting equations in order to obtain the corresponding evolution equations. The latter approach usually requires the functional to be linearized. The second technique, we have been focusing on, is based on first discretizing the functional and minimize the resulting problem using large-scale numerical optimization. The truncated Newton method is an example of a method that only needs to compute at each iteration of the optimization process the function value and its gradient. We have focused our research on both line-search and trust-region approaches embedded in a multigrid technique. The previous technique is currently being applied for optical flow computation. For that issue we have developed a model that is contrast invariant, that is, it is not based on the classical constant luminance hypothesis.

  • Today research is performed at the UB. The research has been focused on different topics such as optical flow methods and parametric active contours. Within this latter topic we have developed the Cage Active Contours; a tool that can be considered a combination of the level sets and the parametric active contours. Currently I do have several topics of interest which are centered on machine leaning. In particular, I am interested in topics such as big data machine learning and data clustering. Contact us if you are interested in these topics!


Lluis Garrido
Department de Matemàtiques i Informàtica
Universitat de Barcelona

C/ Gran Via de les Corts Catalanes, 585
08007 Barcelona

Phone (+34) 934 020 854

E-mail: lluis(.)garrido(at)ub(.)edu

Links @ UB

Compter Vision and Machine Learning Group

Data Science Master

Department de Matemàtiques i Informàtica

Universitat de Barcelona