Well, in the meantime already three nutshells...
Keywords: visual perception, modeling, image processing, pattern recognition,
computer graphics, parallel computing
|
|
| Name | Position | Project/Subject |
| Prof.dr. Hans du Buf | Head of Laboratory | |
| Prof.dr. Hamid Shahbazkia | Ex-Postdoc :-) | ADIAC |
| Eddy Loke | PhD student | ISACS |
| João Rodrigues | PhD student | MOVIDE |
| Roberto Lam | PhD student | Surface rendering - mesh reduction |
| Samuel Nunes | PhD student | Exocet |
| Daniel de Almeida | PhD student | Exocet |
| Name | Grad Student | Direct Volume Rendering |
Previous people:
Postdocs: Jos van Deemter, Stavri Nikolov, Florin Rotaru, Don Gallogly, Ulli Bobinger
PhD students: Jeroen Melchiors, Adrian Ciobanu, Luis Santos
MSc students: João Rodrigues, Roberto Lam, Pedro Guerreiro, Ulrich Schnier
Recent Grads: Ricardo Oliveira, Pedro Semeano, Luis Sismeiro, Filipe Marreiros,
Paulo Oliveira
Others: Thomas Krause
Must be updated, all Silicon Graphics has been substituted...
1 SGI Origin 200QC server with 4 R10000 (each with 2MB L2 cache) and CrayLink
1 Dual AMD-MP server, Linux
1 Dual-PIIIE@750 coppermine with 1 GByte memory, Linux
1 AMD K7 250 GByte RAID fileserver, Linux (*)
2 Dual-PentiumIII@450 PCs with 100Mb/s link and MPI (***)
1 SGI Indigo2/SolidImpact/R10000 station with CristalEyes
4 SGI Indy stations (being replaced by 4 AMD XPs)
N Iomega Jaz drives, Zip drives, 4GB DAT, CDROM R/RW, DVD R/RW
1 Airis Centrino laptop with 60 GB, DVD R/RW and 15.2" screen
6 Pentium PCs running Linux
2 Pentium PCs running Win#&%@$grrr
1 Olympus BX40 microscope
1 Polaroid Digital Microscope Camera (max 1200x1600 in 24 bit color)
1 Physik Instrumente P-721.10 piezotranslator (max 90 microns; 10nm steps)
1 Cambridge VSG2/3F Visual Stimulus Generator with OptiCal
1 HP ScanJet 2400 dpi scanner
1 Lexmark 1200x1200 dpi b/w laser printer
1 HP 2300 b/w laser printer
1 Epson Photo 750 colour printer
9 APC UPS systems
PlayStation2 running Linux - read ps2.txt.
(*) Backups and the fileserver: read void.txt.
(***) The two dual-Pentium cluster named Ping-Pong was replaced by the "Calhau", an 8-node AMD XP Beowulf cluster, but Ping is the front-end of the cluster. See CALHAU.
See this ASCII file for some benchmarking...
For new users on the Origin 200: please take a look at the ASCII file o200 for getting started (compilation, optimization, parallelization, program timing, etc, etc). File updated Sept 7, 1998
All equipment has been acquired by project funding.
The Cambridge Visual Stimulus Generator is for doing
psychophysical experiments (AMOVIP project). The microscope with CCD and
piezotranslator is for autofocusing and optical sectioning microscopy,
eg 3D reconstruction and visualization (ADIAC project).
The lab can also use a Compaq/DEC 4100 server with four Alpha processors at 466 MHz.
For a complete list of books and proceedings etc in the lab, go to our LIBRARY
SmartVision - Active vision for the blind - Started Jan 2008; three years; webpage almost up to date...
EXOCET/D started in 2004: Extreme Ecosystem Studies in the Deep Ocean - Technological Developments. 6th FP Global Change and Ecosystems. Duration 3 years.
SPMDlib - High-level parallelization tools
See also Introduction and Directive
design
This project concerns the development of tools for a semi-automatic
program parallelization. At the moment we have a very easy fortran
library (C in development) to split loops and to send parts of or entire arrays.
This hides all MPI stuff from a user and enables a very fast parallelization
with a minimum programmer intervention. We also develop a set of
compiler directives (a la Silicon Graphics' C$DOACROSS or OpenMP) plus a
preprocessor or parser that puts all necessary spmd_ calls into the source
program such that it can be compiled and executed on a Beowulf.
An important aspect will be to estimate the speedup and to optimize the number of
processors used. This project was/is part of PARTOOLS
Please see the detailed project homepage!
MOVIDE - Modeling Visual Detection
Duration: Oct 2000 - Oct 2003
Funding: FCT/MCT Lisbon
The MOVIDE project webpage is under construction. The goal of MOVIDE is to develop models that can predict threshold data for visual detection. Is a collaboration with Uwe Mortensen and Guenter Meinhardt of Univ. of Muenster.
ISACS - Integrated System for the Analysis and
Characterization of the Seafloor
Duration: March 1996 - March 1999; extension until Sept 99
Funding: European Commission, MAST3 programme
This project concerns the analysis of the seabottom in three dimensions using
sonar. Our task is to develop 3D segmentation and visualization techniques.
Please see the project homepage for further details like simulations, results,
and papers.
This project got in December 1997 additional support in terms of a NATO
Collaborative Research Grant (CRG 971584) for The Analysis of Acoustic
Data via 3-D Segmentation with Todd Reed (see affiliations below).
Please see the detailed project homepage!
AMOVIP - Advanced Modeling
of Visual Information Processing
Duration: April 1998 - April 2001
Funding: European Commission, INCO-ESPRIT programme
This project concerns: (A) The modeling of spatial brightness perception by
means of Gabor filters in combination with multiscale line/edge and
vertex detection with a scale-stability analysis. This symbolic
representation will be used to construct brightness maps which will
show the known brightness effects such as Mach bands, induction and
assimilation. For literature see du Buf's publication list
(papers that appeared in Optical Engineering and Biological Cybernetics).
(B) The application of the model components in image denoising and
coding, but applying wavelets and Hermite polynomials.
See the detailed project homepage!
ADIAC - Automatic Diatom Identification And
Classification
Duration: May 1998 - May 2001
Funding: European Commission, MAST3 programme
This project is a pilot study concerning the automisation of diatom
identification on the basis of images. Diatoms are unicellular algae and
the frustules have a very regular ornamentation and shape.
If you have never seen them, you can go to the ADIAC pubdat pages.
It is planned to apply different analysis schemes (classical thresholding
and thinning, mathematical morphology, multiscale line/edge extraction
using Gabor filtering) together with classification (eg graph matching).
Other tasks in the project are autofocusing and estabishing a huge
database with images.
Coordinator: Hans du Buf.
Please see the detailed project homepage!
SETNA-ParComp
- Scalable Environments and Parallel Computing
Duration: August 1997 - November 2000
Funding: PRAXIS XXI, TIT programme
This project concerns the development of tools for parallel computing,
numerical algorithms, and computer vision. Our task is to develop
scalable 3D segmentation methods on distributed memory systems (Parsytec CC).
This project is best characterized as financial poverty...
Texture, multiscale lines/edges, visual perception (ALSO NEW PROJECTS)
Duration: January 1995 - February 1996
Funding: European Commission, HCM research network
This project concerned multiscale Gabor/wavelet analysis in which the local phase is used for line/edge classification. A publication is in preparation, but this will not include texture (Jos van Deemter). This work will be continued in one of the new projects (AMOVIP), see HERE for ASCII text concerning new project positions (right now one postdoc for the ADIAC project). You should also look at this text if you would like to ask for a GRANT (PhD student or postdoc) and work on texture or visual perception here at the Vision Laboratory. For publications on texture and visual perception, look at the dubuf homepage.
Pedro Semeano and Luis Sismeiro finished their final project (Dec 2001). Subject: IP-over-SCSI. We had the first Beowulf cluster on which this really worked! See IP-over-SCSI
Ricardo Oliveira also finished his final project (Oct 2001). Subject: C routines on top of MPI for image filtering.
Roberto Lam and Pedro Guerreiro got their Masters' degree in July 2001.
Parabens! Automatically promoted to PhD students :-)
Roberto's thesis is about the smoothing and reduction of triangle meshes for
surface rendering. Pedro's is about simple visual detection models.
Filipe Marreiros started his final-year project on Realtime Direct Volume Rendering, July 2001.
The new coppermine and Beowulf cluster are up and running; see CALHAU and benchmarking.
VOID, our new fileserver: read void.txt.
New books are still arriving. See our LIBRARY
Ulrich Schnier did his Masters' project from September 2000 to April 2001. He came from Klaus Toennies' group at the University of Magdeburg.
New papers:
Complete list: publications.
Two Portuguese projects have been approved in July 2000, duration 3 years:
MOVIDE - Modeling Visual Detection (one postdoc for the Vision Lab)
and
3D MODELING FROM VIDEO (in collaboration with João Paulo Costeira of the
ISR/Lisbon).
Prof Todd Reed (see Links) visited us in June 99 and gave a two-day introduction to image and sequence coding in our Masters' course.
Dr Guenter Meinhardt from Uwe Mortensen's group (see Links) has visited the Vision Laboratory in May 99. He gave the seminar "Psychophysical evidence for different kinds of adaptivity in the human visual system." It looks like we are going to establish a really active collaboration concerning experiments and Gabor-based modeling of detection thresholds!
Visits:
Since 1997 the Vision Laboratory has been visited by:
Dr. Rolf Wuertz (Center High Performance Computing, Univ of Groningen; he's now
back to the Institut fuer Neuro-Informatik in Bochum)
Prof. Nikolay Petkov (Center High Performance Computing, Univ of Groningen)
Dr. Jos van Deemter (Center High Performance Computing, Univ of Groningen)
Dr. Gabriel Cristobal, see Links and affiliations
Prof. Uwe Mortensen (Universitaet Muenster), who gave a seminar
Neural networks, self organisation and matched filters in the visual
system
Prof. Hans G. Feichtinger (NUHAG, University of Vienna), who gave a seminar
Gabor analysis using general time-frequency lattices
Dr. George Campbell (after learning Japanese at NEC he's now at Equator Technologies
in California)
Dr. Guenter Meinhardt from Uwe Mortensen's group, who gave the seminar Psychophysical
evidence for different kinds of adaptivity in the human visual system
Prof. Todd Reed (and Nancy) from UC at Davis (see Links)
Dr. Jose Pech-Pacheco and Matthias Keil (Instituto de Optica, CSIC Madrid)
and
Prof. Mike Spann and Dr. Jonathan Turner (University of Birmingham, UK).
Complete publication list: PUBLICATIONS.
Student courses: see ICG
For STUDENTS I put information on ftp.ualg.pt for anonymous access:
/users/dubuf/imags contains images of 64k, raw without header
/users/dubuf/pic2ps contains program to convert raw 64k images to postscript
/users/dubuf/unix4dos contains unix commands for DOS, although installing LINUX
is a much better alternative, and not only because of the price...
If you want to view raw images with xv, for example, you can edit an ascii file called
"head" that contains the following four lines:
P5
# comment
256 256
255
Then, assuming you have a 64k raw image with filename "test", do
"cat head test > test.pbm" and you can do "xv test.pbm".
IPTool is a Delphi-based environment for teaching and working with image processing. You can easily create filters in the spatial and frequency domain, apply zooming and many other algorithms. To download (for Windows) click IPTool
TEXTURES can be found here.
LINE AND EDGE DETECTION can be found here.
An absolute must is to visit the
Computer Vision Home Page and the
Vision Science webpages,
where you can find almost all (?) our colleagues in the world working on computer
and/or human vision. Another useful site with lots of information is
CVonline.
You can also take a look at
Wavelets for Texture Analysis,
Wavelets Internet Sources,
and last but not least the
Gabor Digest
as well as the
Wavelet Digest.
If you are interested in neuroscience, including visual psychophysics,
the modeling of visual perception, as well as neurophysiology of cortical
processing, you can use a very powerful search
engine at the Neuroscience
Web Search site.
For discussions, FAQs, etc, see the newsgroups sci.image.processing and
comp.ai.vision. Another useful source of information is Adrian Clark's
PIXEL.
Affiliations:
The Vision Laboratory is affiliated with:
the ISR Signal Processing Lab at the Tecnico in Lisbon.
The Portuguese Association for Pattern Recognition APRP.
Hans du Buf is also a member of the IEEE Computer Society and Associate Editor of the Int J of Pattern Recognition and Artificial Intelligence: IJPRAI.
Apart from the partners in the ISACS, ADIAC, AMOVIP and SETNA-ParComp, projects, the Vision Laboratory is affiliated with the following people:
Prof.dr. Uwe Mortenson
University of Westfalen, Muenster.
This collaboration concerns the modeling of visual perception by
means of matched filters and Gabor filters.
Prof.dr. Todd Reed (trreed@ucdavis.edu)
Department of Electrical and Computer Engineering, University of California, Davis, USA.
The lab has now a NATO Collaborative Research Grant with Todd on 3D segmentation.
This collaboration concerns 3D segmentation by means of Gibbs-Markov random
fields, initially for image sequence coding but in the future for 3D data
processing. If you are interested in Gabor/wavelet/etc filtering, visit
the ISPG homepage and see some very nice demonstrations under "Derivative of
Gaussian Transform." Also: nice demos of coding.
Prof.dr. Edwin Hancock (erh@minster.york.ac.uk)
Department of Computer Science, University of York (UK).
This collaboration concerns stochastic relaxation labeling for
3D segmentation and border detection.
Prof.dr. Mike Spann (M.Spann@bham.ac.uk) (link may be down!)
School of Electrical and Electronic Engineering, University of
Birmingham (UK).
This collaboration concerns graph hierarchical-based segmentation using
robust statistics as well as processing of seismic data.
Dr. Klaus D. Toennies
(klaus@isgnw.cs.uni-magdeburg.de)
Institut fuer Simulation und Grafik (ISG), Universitaet Magdeburg, Germany.
This collaboration concerns 3D scientific visualisation strategies and
computer assisted analysis of medical images.
Dr. Gabriel Cristobal (gabriel@optica.csic.es)
Instituto de Optica (CSIC), Madrid, Spain.
This collaboration concerns multiscale Gabor/wavelet decompositions
for image coding, line/edge detection, etc. He is a partner in the ADIAC and AMOVIP
projects (he's the Coordinator of AMOVIP).
Dr. Josef Bigun (new address in Sweden; to be included, plus link)
This collaboration is a continuation of our common work on texture
feature extraction based on Gabor filtering and segmentation.
Dept of Electronics and Computer Science - FCT University of Algarve Campus de Gambelas 8000 Faro Portugal Tel (+351) 289 800900 extension 7761; lab: 7751 Fax (+351) 289 818560 Prof.dr.ir. J.M.H. du Buf Email: dubuf@ualg.pt Homepage: dubuf
Two DOC files (sorry, we will make ps versions soon) with maps: gambelas.doc, which shows the region around the airport and the Campus de Gambelas, including HdB's home next to the airport and telephone number for the case that all else fails, and campus.doc, showing the Campus with buildings and the location of the laboratory.
University of Algarve
Faculty of Sciences and Technology
CINTAL - Centre for Technological
Research of the Algarve
Don't forget Hein's Law:
Problems worthy of attack prove their worth by hitting back
This page has been visited
times since August 15, 1997.
Last update: Oct 2009 (NOT COMPLETE!) - HdB