[Jozsef Vass]

Jozsef Vass' Home Page

Contact Information

Jozsef Vass
201 Engineering Building West
Department of Computer Engineering and Computer Science
University of Missouri-Columbia
Columbia, MO 65211
USA
Tel: (573) 884-6049 (office), (573) 882-6265 (lab)
Fax: (573) 882-8318
E-mail: vass@cecs.missouri.edu

I am currently looking for employment opportunities. My objective is to become a member of technical staff of software development in the areas of multimedia networking, audio/image/video streaming, wireless communications, digital signal processing, and standardization.

Education

    Ph.D., Computer Engineering and Computer Science
      University of Missouri-Columbia, Jan. 2000
        GPA: 4.0/4.0
        Thesis: Advanced Wavelet Image and Video Coding Strategies for Multimedia Communications

    M.Sc., Electrical and Computer Engineering
      University of Missouri-Columbia, Dec. 1996
        GPA: 4.0/4.0
        Thesis: Toll Quality Low Bit Rate Speech Coding and Morpho-Subband Image Coding

    Dipl. in Electrical Engineering
      Technical University of Budapest, June, 1995
        GPA: 4.63/5.0
        Thesis: Robustness of Feature Extracted Parameters in Speech Processing (in Hungarian)

Experience

9/95-present Graduate Research Assistant, Multimedia Communications and Visualization Laboratory, Department of Computer Engineering and Computer Science, University of Missouri-Columbia, Columbia, MO
Conducted in-depth research and development on a wide range of projects.  Research results were published in more than 35 technical journal and conference papers.  Proficient in speech (G.711, G.728, GSM, etc.), image (JPEG), and video (MPEG and H.26x) compression standards. Familiar with Internet and ATM technology. Extensively  participated in preparation of funding proposal to NSF, NASA, NIH, and DARPA.

In addition to my research, I am also the system administrator of the Multimedia Communications and Visualization Laboratory (MCVL). I manage 15 SGI workstations (1 Onyx2, 4 Octanes, 8 O2s, and 2 Indigo2s) and several PCs running Windows 98/NT. This work includes both hardware and software support, replacing defected hardware, installing patches,  licensing softwares, etc.
6/96-8/96 Summer Internship, NASA Goddard Space Flight Center, Greenbelt, MD
Developed and implemented a novel image matching algorithm by using relaxation labeling for automatic cloud height estimation on massively parallel computer (MasPar).  A grant proposal to NASA and several papers are resulted from this three month visit.
9/93-6/94 Undergraduate Research Assistant, Department of Telecommunications and Telematics, Technical University of Budapest, Hungary
Participated in the development of a neural network-based approach for traffic optimization in ATM networks. (Funded by Ellemtel Laboratories (Ericsson)). A patent and several publications are resulted from this ten month work.

Computer Skills

Environments: Several flavors of UNIX (SunOS, Solaris, IRIX, and AIX), Windows 95/98/NT, Macintosh, Next, and IBM mainframe (VM/CMS)
Languages: C, C++, Java, Pascal, Basic, Assembly (8086 and TMS320C30), HTML, and scripting languages (csh, tcsh, and Perl)
Softwares: MatLab, Microsoft Office, WordPerfect, Mathematica, Khoros, Adobe PhotoShop, Oracle, TeX, and LaTeX
Develpment Tools: OpenGL, X Windows/Motif, CosmoCreate

Research Projects

Video Streaming over Wireless and Wireline Networks: For the wireless case, we have developed a proxy-based system, where the video is adapted to both the hardware capabilities of mobile hosts and time-varying wireless channel conditions. At lower network layers, we use either the ITU-T H.223 Recommendation or wireless ATM.  For video compression, we use our  three-dimensional significance-linked connected component analysis (3D-SLCCA).  Real-time handling of wireless transmission errors is carried out by using our multilayer error protection strategy, which includes hierarchical resynchronization, unequal forward error protection, and retransmission as well.  For the wireline case, we are studying scalable video transmission by using 3D-SLCCA source codec over multicast IP networks. Channel Coding: The previously developed significance-linked connected component analysis (SLCCA) image codec is enhanced for image transmission over noisy channels.  The developed multilayer error protection strategy includes error resilient packetization, hierarchical resynchronization, and unequal error protection. Wavelet-based Video Compression: Two algorithms were developed for video communications, namely, video significance-linked connected component analysis (VSLCCA) and three-dimensional significance-linked connected component analysis (3D-SLCCA). In VSLCCA, fine-tuned motion estimation and exhaustive overlapped block motion compensation ensures blocking-effect-free error frames and thus wavelet transform can be efficiently applied.  Significant wavelet coefficients are represented by using the SLCCA technique. VSLCCA provides superior performance to H.263. VSLCCA was also implemented as a JAVA applet and a standalone application using Motif. (More information.)

In 3D-SLCCA, both the wavelet transformation and SLCCA data organization and representation strategy is extended to 3-D to include the time domain. In addition to high coding efficiency, 3D-SLCCA provides scalability, low computational complexity, and error resilience. 3D-SLCCA outperforms 3D-SPIHT.  At high bit rate, 3D-SLCCA provides superior image quality when compared to MPEG-2.

Wavelet-based Image Compression: We have developed a high performance image coding algorithm termed significance-linked connected component analysis (SLCCA).  SLCCA exploits both the within-subband clustering property and cross-scale similarity of wavelet transformed images.  As a result, SLCCA outperforms Shapiro's embedded zerotree wavelet (EZW), Servetto et al. 's morphological representation of wavelet data (MRWD), and Said and Pearlman's set partitioning in hierarchical trees (SPIHT) as well.  (More information) Speech Coding:  We have developed a high performance linear predictive coding (LPC) quantization scheme that was integrated with the  FS1016 code excited linear predictive (CELP) speech codec.  Superior results were obtained when compared to the original scalar LPC quantizer. Image Database Retrieval:  We have developed an effective and efficient image database system by using a unified image compression algorithm for image storage, indexing, and retrieval.  As a result, the complexity and storage requirement of the database management system  are significantly reduced.  In this research, we have chosen SLCCA as the underlying compression algorithm, and features used for indexing are directly extracted from the SLCCA bitstream. The main features of the developed system include 1) progressive image transmission, 2) interactive search refinement, and 2) dynamic scalable features. (More information)   Morpho-Subband Decomposition: The developed low bit rate morpho-subband image coding scheme adaptively combines linear and nonlinear subband decomposition.  For uniform and texture regions, linear subband decomposition is used. For edge regions, morphological decomposition is used.  The developed algorithm outperformed Egger et al. 's adaptive subband decomposition (ASD) (O. Egger, W. Li, and M. Kunt, "High compression image coding using an adaptive morphological subband decomposition," Proc. IEEE, vol. 83, no. 2, pp. 272-287, Feb. 1995) by 2.71 dB in peak signal-to-noise ratio (PSNR) for the Lena image at 0.16 bits-per-pixel. Object-based Video Coding: The emerging MPEG-4 standard composes the scene of meaningful audio-visual objects. In the case of natural sequences, the main problem is image segmentation. The proposed technique combines temporal and spatial information to obtain semantically meaningful segmentation of moving video objects. Land Cover and Land Use Automatic Classification: Currently, land cover and land use classification for the state of Missouri is semi-automatically carried out by MoRAP.  We are developing automatic algorithms that can significantly speed up this process. Our research directions include 1) binary tree genetic algorithm, 2) Bayesian classification with learning, 3) and the application of Gaussian mixture density modeling. Other Research Projects: 3-D graphics and topology coding; digital image watermarking; robust estimation; relaxation labeling, and cloud height estimation from stero images.

Patent

  1. M. Boda, T. Szecsy, S. Blaabjerg, J. Biro, J. Vass, T. Tron, and A. Farago, "Device and method for distribution of resources of a physical network," filed March 8, 1995, foreign application no. 9500838-9 (patent right: Ellemtel UA, Sweden), granted U.S. patent no. 5687292, Nov. 11, 1997.

Publications

During my Ph.D. study, I have published more than 35 technical journal and conference papers.   Please see the  complete list of my publications.
 

Honors and Activities

  IEEE Student  Member
   Member of Reviewer for Participated in the Research and Creative Activities Forum at the University of Missouri-Columbia in 1997 and 1998. (Awarded second place in 1998.)
Studied a semester at the University of Missouri-Columbia as an exchange student (Fall 1994)
Awarded "The Scholarship of the Hungarian Republic"
Received "Scholarship for the Hungarian Technical Progress"
Participated in a short course on "Digital Signal Processing" at Technical University of Delft, Delft, The Netherlands, March 1993

Coursework at the University of Missouri-Columbia

[Home] CECS Multimedia Visualization and Communications Laboratory

Last  revised: Nov. 4, 1999