research

People

Current Students Catarina Barata (PhD) Instituto Superior Técnico Carlos Santiago (PhD*) Instituto Superior Técnico *co-supervised.

Former Students Arnaldo J. Abrantes (MSc, PhD) Polytechnic Institute of Lisbon Jacinto Nascimento (MSc, PhD,PD) Researcher at ISR João Sanches (PhD) Instituto Superior Técnico Pedro M. Jorge (MSc, PhD*) Polytechnic Institute of Lisbon Jorge G. Silva (MSc, PhD*) Duke University, USA Lourenço Bandeira (PhD*) Instituto Superior Técnico Jorge F. Nunes (MSc) Pedro Sequeira (MSc) Rui Jesus (MSc*) Polytechnic Institute of Lisbon Fábio Silva (MSc*) Ricardo Martins (MSc*) Luis Rosado (MSc*) João Pacheco (MSc*) João Fernandes (MSc*) Mitsubishi Europe Carlos Santiago (MSc*) Instituto Superior Técnico Catarina Barata (MSc*) Instituto Superior Técnico Eduardo Bicacro (MSc*) Instituto Superior Técnico Sérgio Copeto (MSc*) Instituto Superior Técnico Ana Mazo (MSc*) Instituto Superior Técnico Marta de Aguero (MSc*) Instituto Superior Técnico Mariana Francisco (MSc*) Instituto Superior Técnico Margarida Ruela (MSc*) Instituto Superior Técnico Pedro Morgado (MSc*) Instituto Superior Técnico *co-supervised.

Collaborations André Marçal University of Porto Andrew Gee University of Cambridge Fausto Pinto Hospital de Santa Maria Gilles Celeux Université Paris-Sud João Miranda Lemos Instituto Superior Técnico Jorge Rozeira Hospital Pedro Hispano Margarida Silveira Instituto Superior Técnico Mário Figueiredo Instituto Superior Técnico Miguel Barão University of Évora Nicolás Pérez de la Blanca University of Granada Paula Rocha University of Porto Paulo Jorge Ferreira University of Aveiro Pedro Pina Instituto Superior Técnico Richard Prager University of Cambridge Teresa Mendonça Universidade do Porto

Topics

I work in image and signal processing from a model based point of view. The main question I try to address is: how can we extract useful information from signals and images and e.g., segment organs in medical images, track pedestrians in video signals or detect abnormal behaviors.

I have been involved in the following problems:

Image Processing

(object tracking and robust deformable models)

I am interested in the estimation and tracking of moving objects in video sequences, using deformable models (elastic curves and surfaces). The main challenges I try to address concern the development of robust estimation methods for shape tracking.

Refs: 1, 2, 3, 4

This work was done in collaboration with Jacinto Nascimento, Arnaldo Abrantes, Gilles Celeux, João Lemos, Margarida Silveira.

Biomedical Applications

I have been involved in two problems: estimation of 3D volumes from a set of noisy ultrasound images (3D ultrasound) and organs segmentation and tracking using robust deformable models.

We developed volume reconstruction algorithms for free hand 3D ultrasound. Volume reconstruction is done taking into account multiplicative (speckle) noise, geometric deformations and nonlinear measurements which occur duting the acquisition process.

We have also developed robust tracking algorithms for deformable shapes with outlier rejection and multiple dynamic models. We have applied these algorithms to track the left ventricle of the heart in ultrasound images.

We are currently involved in the project ADDI.

Refs: 5, 6, 7, 8

This work was done in collaboration with João Sanches, Jacinto Nascimento.

Surveillance

This work aims to track moving pedestrians in a vídeo-surveillance system and to characterize their activities.

We have developed a system to track pedestrians in groups which can deal with group merging and splitting. This system is based on statistical inference with Bayesian networks.

We can also characterize human activities using hierarchical dynamical models. We learn a set of atomic activities from the video sequence (standing, moving left, right, etc) are define higher order models based on sequences of atomic activities.

We are currently involved in the project ARGUS.

Refs: 9, 10, 11

This work was done in collaboration with Jacinto Nascimento, Mário Figueiredo, Arnaldo Abrantes, Pedro Jorge, João Lemos,

Manifold Learning

The image of a face has hundreds of variables. The space dimension n is very high. However there is a strong dependence among these variables. Valid face images belong to a subset (manifold) of the n-dimensional space with a low intrinsic dimension. How can we estimate this subset (manifold), its dimension and use this information to improve image processing task? This problem has been recently addressed by a number of researchers.

We tried to address these questions in the Ph. D. work of Jorge Silva and develop a method for manifold estimation which we call tangent bundle approximation.

Refs: 12, 13

This work was done in collaboration with Jorge Silva, João Lemos.

Selected Publications

1. Arnaldo J. Abrantes, A Class of Constrained Clustering Algorithms for Object Boundary Extraction, IEEE Trans. Image Processing, 1507-1521, 1996.

2. Gilles Celeux, Jorge S. Marques, Jacinto Nascimento, Learning Switching Dynamic Models for Objects Tracking, Pattern Recognition, vol. 37, nº9 , 1835-1840, September 2004.

3. Jacinto Nascimento, Jorge S. Marques, Robust Shape Tracking in the Presence of Cluttered Background, IEEE Transactions on Multimedia, Vol. 6, no. 6, December, 2004.

4. Jacinto C. Nascimento, Jorge S. Marques, Adaptive Snakes Using the EM Algorithm, IEEE Transactions on Image Processing, Volume 14,  Issue 11,  1678 – 1686, Nov. 2005.

5. João M. Sanches, Jorge S. Marques, Joint image registration and volume reconstruction for 3D ultrasound, Pattern Recognition Letters, Special Issue on Ultrasonic Image Processing and Analysis, vol 24, 791-800, February 2003.

6. João M. Sanches, Jorge S. Marques, A Multi-Scale Algorithm for Three Dimensional Free Hand Ultrasound, Ultrasound in Medicine and Biology, Vol. 28, 1029-1040, August 2002.

7. Jacinto Nascimento, João Sanches, Jorge S. Marques, A Method for the Dynamic Analysis of the Heart Using a Lyapounov Based Denoising Algorithm, IEEE Engineering in Medicine and Biology Society (EMBS), New York, USA, Aug., 2006.

8. Jacinto Nascimento, João M. Sanches, Jorge S. Marques, Tracking the left ventricle in ultrasound images based on total variation denoising, Pattern Recognition and Image Analysis, LNCS 4477, vol. 2, 628-635, June 2007.

9. Jacinto C. Nascimento, Mario A. T. Figueiredo, Jorge S. Marques, Independent Increment Processes for Human Motion Recognition, Computer Vision and Image Understanding, accepted, 2007.

10. Jacinto C. Nascimento, Jorge S. Marques, Performance Evaluation of Object Detection Algorithms for Video Surveillance, IEEE Transactions on Multimedia, Vol. 8, No. 4, 761-774, August 2006.

11. Pedro M. Jorge, Arnaldo J. Abrantes, J. M. Lemos, Jorge S. Marques, Long Term Tracking of Pedestrians with Groups and Occlusions, in Bayesian Network Technologies: Applications and Graphical Models, Ankush Mittal and Dr. Ashraf Kassim (eds.), Chapter VIII, in press, 2007.

12. Jorge G. Silva, Jorge S. Marques, João M. Lemos, Selecting Landmarkpoints for Sparse manifold Learning, NIPS, 2005.