Curriculum vitae

Name:

Dr Zoltán KATÓ

Born:

February 9, 1965, Békéscsaba, Hungary

Citizenship:

Hungarian

STUDIES:

1991-94

Graduate student at INRIA - Sophia Antipolis (The French National Institute for Research in Computer Science and Control), France.
PhD with honors in Computer Sciences.

1992

NATO Summer school on ``Progress in Image Processing'', Les Houches, France.

1990-91

DEA (Diploma of special studies) ``Vision and Robotic'' at the University of Nice, France.

1990

Master of Science in Mathematics and Software Engineering.

1988

Bachelor of Science in Mathematics and Software Engineering.

1985-90

Full time student at the József Attila University, Szeged, Hungary.

LANGUAGES:

English, French.

INTERESTS:

image processing, motion, color, texture, segmentation, multi-scale methods, multi-media, content based indexing and retrieval in multimedia databases, Markov random field modelization, parameter estimation, parallel algorithms, optimization algorithms, stochastic relaxation, GIS (Geographical Information Systems), software development.

GRANTS:

2003-2006

Bolyai Janos Research Fellowship of the Hungarian Academy of Sciences.

2000-2002

My research proposal Stochastic Paintbrush Image Transformation has been granted by the School of Computing, National University of Singapore. This is a collaborative project with the research group of Tamas Sziranyi at the Image Processing and Neurocomputing Department of the University of Veszprem, Hungary.

1998

Our research proposal on color image segmentation with T. C. Pong and C. M. Lee on color image segmentation, has been granted by the Hong Kong Research Grants Council.

1997-98

18 months ERCIM (European Research Consortium for Informatics and Mathematics) postdoctoral fellowship.

1995

A joint INRIA and SZTAKI (Computer and Automation Inst., Hungarian Academy of Sciences) research project, based on the results of my PhD work, has been granted for 1 year by the Balaton program of French and Hungarian Government.

1990-94

4 years PhD Fellowship of the French Government.

SCIENTIFIC APPOINTMENTS / COLLABORATIONS:

2002-

Assistant Professor at Department of Computer Science, Szeged University, Hungary

2000-2001

Fellow at School of Computing, National University of Singapore.

1999

Research associate at The Hong Kong University of Science and Technology, Computer Science Department, Hong Kong (T. C. Pong)

1997-98

ERCIM postdoctoral fellowship at Signals and Images group, CWI (National Research Institute for Mathematics and Computer Science in the Netherlands), Amsterdam (18 months).

1996-97

Research associate at The Hong Kong University of Science and Technology, Computer Science Department, Hong Kong (1 year with T. C. Pong and C. M. Lee).

1995

Joint research collaboration between INRIA and SZTAKI (Computer and Automation Inst., Hungarian Academy of Sciences) founded by the French and Hungarian Government. I was working as an external fellow with Tamas Szirányi (Analogical and Neural Computing Systems Lab., SZTAKI), Josiane Zerubia and David Geldreich (PASTIS group, INRIA).

1995

3 months research engineer appointment at INRIA-Sophia Antipolis, France.

1991-94

PhD thesis at PASTIS group, INRIA-Sophia Antipolis, France (4 years with Josiane Zerubia and Marc Berthod). During my PhD, I have done some research work for the CNES (French Space Agency).

REFEREEING:

I have refereed papers submitted to IEEE Transactions on Signal Processing Letters, IEEE Transactions on Image Processing, Pattern Recognition Letters, Image and Vision Computing, International Conference on Pattern Recognition (2002), IEEE International Conference on Intelligent Robots and Systems (1999), IEEE Asian Conference on Computer Vision (1998, 2002). I also served as a session chair of the "Segmentation and Grouping" session at ACCV'98.

TEACHING:

Computer Graphics, Digital Image Segmentation, Operating Systems, Computer Organization, Programming in UNIX, Data Structures and Algorithms.

RESEARCH PROJECTS:

2000-2002:

Stochastic Paintbrush Image Transformation Stochastic paintbrush rendering is an algorithm to simulate the painting process to get a picture similar to a real painting, where the purpose of the painter is to portray something which looks like to be a real scenery. The overall goal of this project is to investigate both theoretical and applied aspects of this new algorithm. The proposed research directions include: image similarity measure used in an image indexing and retrieval system and image segmentation.

1999:

Object Segmentation in Video Sequences for MPEG-4 Using a Combination of Color and Motion Information We started this project in January with T. C. Pong at HKUST. The goal is to develop a model which is able to combine different features (in the first step color and motion) and detect objects in a video sequence for MPEG-4 encoding.

1997-98

Unsupervised Image Segmentation for Automatic Image Indexing and Retrieval Using Color and Spatial Features During my ERCIM fellowship at CWI, I was working on the problem of detecting meaningful, homogeneous regions in an image based on color. Such a preprocessed image can then be used for automatic indexing in a large database of images using color and shape information. The goal of this project is to propose a method which is able to segment a color image without any human intervention. The only input is the observed image, all other parameters are estimated during the segmentation process. Our method is model-based, we use a first order Markov random field (MRF) model (also known as the Potts model) where the singleton energies derive from a multivariate Gaussian distribution and second order potentials favor similar classes in neighboring pixels. The most difficult part is the estimation of the number of pixel classes or in other words, the estimation of the number of Gaussian mixture components. Reversible jump Markov chain Monte Carlo (MCMC) is used to solve this problem. These jumps enable the possible splitting and merging of classes. The algorithm finds the most likely number of classes, their associated model parameters and generates a segmentation of the image by classifying the pixels into these classes. The estimation is done according to the Maximum A Posteriori (MAP) criteria. Experimental results are promising, we have obtained accurate results on a variety of real color images.

Besides working on this topic, I have participated in the preparation of a research proposal on content based indexing and retrieval in large image databases with Henk Heijmans, Fons Kuijk, Rob Zuidwijk and Ben Schouten.

1996-97

Motion Compensated Color Image Segmentation and Parameter Estimation in a Markovian Framework This project is about unsupervised, motion compensated color image segmentation algorithms. The model is based on an earlier intensity based MRF model proposed in my PhD thesis. In the new model, we use the CIELUV color metric because it is close to human perception when computing color differences. On the other hand, intensity and chroma informations are separated in this space. First, a still image segmentation model is considered, then we extended our model to take into account motion information when working on color video sequences. We also developed a new parameter estimation method to estimate the mean vectors effectively even if the observed image is noisy and the histogram doesn't have clearly distinguishable peaks. These values are then used in a more complex, iterative estimation process as initial values. The methods have been tested on a variety of real images (indoor, outdoor), noisy video sequences and noisy synthetic images. I worked on this project with T. C. Pong, C. M. Lee and other fellows working on the VIDEOBOOK project at HKUST.

1995-96: MRF Image Segmentation on a CNN Chip

In January 1995, a scientific collaboration has begun between INRIA and SZTAKI (Computer and Automation Inst., Hungarian Academy of Sciences) founded by the French and Hungarian Government. The goal is to implement the Markovian image segmentation model proposed in my PhD thesis on a new chip called CNN (Cellular Neural Networks) which is capable to do the task in real time. I have participated in this project as an external fellow giving some advices and explanations on the MRF model. The main drawback of Markovian models is that they result in a complex cost function whose optimization needs a huge amount of computing power. The main contribution of this project is that our algorithm on a real VLSI CNN chip can execute a pseudo-stochastic relaxation algorithm of about 100 iterations in about 1msec. This new technique enables the use of MRF models in applications where real time processing is needed. I have worked on this project with Tamas Szirányi (Analogical and Neural Computing Systems Lab., SZTAKI), Josiane Zerubia and David Geldreich (PASTIS, INRIA).

1991-94

Multi-scale Markovian Modelization in Computer Vision with Applications to Satellite Image Segmentation (PhD thesis) My thesis advisors were Josiane Zerubia, head of the ARIANA group, and Marc Berthod (former head of PASTIS group, currently director of INRIA Sophia Antipolis research unit).

The main concern of my PhD thesis is Markovian modelization in early vision. The following topics have been discussed in detail:

• How to build Markovian image models in early vision.
• Optimization of non-convex energy functions.
• Parameter estimation, data-driven algorithms.

The main contribution is a new hierarchical MRF model and a Multi-Temperature Annealing (MTA) algorithm proposed for the energy minimization of the model. The convergence of the MTA algorithm has been proved towards a global optimum in the most general case, where each clique may have its own local temperature schedule. Our MTA theorem is a generalization of the famous Annealing theorem of S. Geman and D. Geman (in Stochastic relaxation, Gibbs distribution and the Bayesian restoration of images. IEEE-PAMI, 6:721-741, 1984)

PROGRAMMING LANGUAGES:

C, C++, C*, Pascal, PL1, Assembly.

COMPUTERS/OPERATING SYSTEMS:

IBM/PC (DOS, Windows, Linux), Sun (SunOs, Solaris), DEC Alpha (OSF 1), SiliconGraphics (IRIX), Connection Machine CM200.

OTHER SKILLS:

Xwindows, Motif, TCP/IP network programming, PostScript, TeX, LaTeX, MSWord, Xemacs, Maple, Matlab, Oracle, Arc/Info, ArcView, dBase, Clipper, HTML and many other utilities.

PROGRAMMING EXPERIENCE:

10 years programming experience in C on Sun (Unix, Xwindows, Motif), IBM PC and 4 years in C* (a parallel extension of the C language) on a Connection Machine CM200 (a SIMD parallel computer).

THE MOST IMPORTANT INDUSTRIAL PROJECTS:

1990

I have developed one of the most popular dictionary program with Tibor Kokeny at Scriptum Ltd., Hungary.

1991-93

I have implemented an image segmentation program on a parallel SIMD computer (Connection Machine CM200) using three different Markovian models and six different optimization algorithms. This program have been developed at the INRIA in collaboration with the CNES (French Space Agency).

1994

I have developed an image visualization program for INRIA and CNES on a Sun workstation under the X windows system and Motif.

1994

On a Sun workstation, I have implemented a sequential version of the image segmentation program running on the CM200 for the CNES.

1995

I have implemented a parameter estimation program for INRIA and CNES in order to do unsupervised image segmentation.

1995-96:

I was working on a PHARE project (a government project supported by the European Union) as project manager at the Rudas & Karig Ltd., Hungary in collaboration with the Geological Institute of Hungary. Our task was to design and implement a complex geographical information system (GIS) for the Ministry of Environment.

2002:

I was working as a consultant on IBM's MOF (Manufacturing of the Future) project. Our task was to design and implement the image processing modules for the automated inspection of the HDD slider assembly.

PUBLICATIONS:

1. Z. Kato, Ji Xiaowen, T. Sziranyi, Z. Toth, L. Czuni: Content-Based Image Retrieval Using Stochastic Paintbrush Transformation. ICIP, New York, USA, Sep 2002

2. Z. Kato, T. C. Pong, S. G. Qiang: Multicue MRF Image Segmentation: Combining Texture and Color. ICPR, Quebec, Canada, Aug. 2002

3. J. Zerubia, A. Jalobeanu, Z. Kato: Markov Random Fields in Image Processing. Application to Remote Sensing and Astrophysics. Chapter in the book: A. Bijaoui and J.P.Rozelot editors, New avenues for astronomical data analysis. Springer, 2001

4. Z. Kato, T. C. Pong: A Markov Random Field Image Segmentation Model Using Combined Color and Texture Features. In W. Skarbek editor, Proceedings of International Conference on Computer Analysis of Images and Patterns, 547-554, Springer; Warsaw, Poland, Sep. 2001

5. Z. Kato, T. C. Pong, J. C. M. Lee: Color Image Segmentation and Parameter Estimation in a Markovian Framework. Pattern Recognition Letters, Vol. 22, No. 3-4, 309-321, Mar. 2001

6. T Sziranyi, J Zerubia, L Czuni, D Geldreich, Z Kato: Image Segmentation Using Markov Random Field Model in Fully Parallel Cellular Network Architectures. Real Time Imaging, Vol. 6, No 3, 195-211 Jun. 2000

7. Z. Kato, J. Zerubia, M. Berthod: Unsupervised Parallel Image Classification Using Markovian Models. Pattern Recognition, Vol. 32, 591-604, 1999

8. Z. Kato: Bayesian Color Image Segmentation Using Reversible Jump Markov Chain Monte Carlo. Research Report 01/99-R055, ERCIM (European Research Consortium for Informatics and Mathematics), also available as a CWI Research Report PNA-R9902, ISSN 1386-3711, Jan. 1999

9. Z. Kato, T. C. Pong, J. C. M. Lee: Motion Compensated Color Video Classification Using Markov Random Fields. In Proc. ACCV'98, Hong Kong, Jan. 8-11, 1998

10. Z. Kato, T. C. Pong, J. C. M. Lee: Motion Compensated Color Image Classification and Parameter Estimation in a Markovian Framework. Technical Report HKUST-CS97-04, The Hong Kong University of Science and Technology, Hong Kong, July 1997

11. Z. Kato, T. C. Pong, and J. C. M. Lee: Color Image Classification and Parameter Estimation in a Markovian Framework. In H. T. Tsui and R. Chung, editors, Workshop on 3D Computer Vision 97, pages 75-79, The Chinese University of Hong Kong, May 1997.

12. T. Sziranyi, J. Zerubia, D. Geldreich, Z. Kato: Cellular Neural Network for Markov Random Field Image Segmentation. In Proc. CNNA'96, IEEE, Seville, pp.139-144, 1996

13. Z. Kato, J. Zerubia, M. Berthod: Unsupervised Parallel Image Classification Using a Hierarchical Markovian Model. ICCV, Cambridge, MA, June 20-23, 1995.

14. Z. Kato, M. Berthod, J. Zerubia, W. Pieczynski: Unsupervised Adaptive Image Segmentation. ICASSP, May 8-12 1995, Detroit, Michigan, USA.

15. Z. Kato, J. Zerubia, M. Berthod: Unsupervised Parallel Image Classification Using a Hierarchical Markovian Model. Research Report No 2528, INRIA, April 1995.

16. M. Berthod, Z. Kato, S. Yu, J. Zerubia: Bayesian image classification using Markov Random Fields. Image and Vision Computing 14(1996): 285-295, 1996.

17. Z. Kato, M. Berthod, J. Zerubia: A hierarchical Markov Random Field Model and Multi-Temperature Annealing for parallel image classification. In CVGIP Graphical Models and Image Processing, Vol. 58, No 1, January 1996, pp 18-37.

18. J. Zerubia, Z. Kato, M. Berthod: Multi-Temperature Annealing: a new approach for the energy-minimization of hierarchical Markov Random Field models. ICPR-Computer Vision and Applications, Jerusalem, Oct. 1994.

19. M. Berthod, Z. Kato, J. Zerubia: DPA: A Deterministic Approach to the MAP. In IEEE Trans. on Image Processing, Vol. 4, No 9, September 1995.

20. Z. Kato, M. Berthod, J. Zerubia: A hierarchical Markov random Field Model and Multi-Temperature Annealing for parallel image classification. Research Report No 1938, INRIA, Aug. 1993.

21. Z. Kato, M. Berthod, J. Zerubia: Multi-scale Markov Random Field models for parallel image classification. In Proc. ICCV, Berlin, May 1993.

22. Z. Kato, M. Berthod, J. Zerubia: Parallel image classification using multi-scale Markov Random Fields. In Proc. ICASSP, Minneapolis, Apr. 1993.

23. Z. Kato, J. Zerubia, M. Berthod: Bayesian image classification using Markov Random Fields. In A. Mohammad-Djafari and G. Demoments editors, Maximum Entropy and Bayesian Methods, pp375-382, Kluwer Academic Publisher, 1993.

24. Z. Kato, J. Zerubia, M. Berthod: Image classification using Markov Random Fields with two new relaxation methods: Deterministic Pseudo Annealing and Modified Metropolis Dynamics. Research Report No 1606, INRIA, Feb. 1992.

25. Z. Kato, J. Zerubia, M. Berthod: Satellite image classification using a Modified Metropolis Dynamics. In Proc. ICASSP, San-Francisco, California, USA, Mar. 1992.