Head of Sector – Dr. Nikolay Merzlyakov

The leading researchers of the laboratory include:

Algorithms, methods and tools for compact description of watermark contour images have been developed. The methods and algorithms were used for creation of database of medieval watermark images.

Methods for watermark identification and generation of their description have been developed. It is based on dynamic generation of a hierarchical tree-like presentation for all specified watermark classes. The localization of each specific class and associated subclasses is determined by program-generated codes. The method is realized at creation of the prototype of the integrated system for digital images processing and database management. Efficiency of this approach has been shown on an example of construction of software for automation of the watermark descriptions in a relational database of watermarks of historical documents.

The database of a relational type, which is offered for using as a typical structure for representation of the archival information stored in funds of the Archive of the Russian Academy of Science, has been developed. Developed structure is focused on the funds containing mainly graphic information requiring authentic presentation (a photo, figures, drawings, hand-written documents). Besides graphic images, fields of the developed database contain additional information for describing fund or the category of archive, and also concrete units of storage (the inventory, the summary). In addition, the fields of the attributes are inserted assisting to carry out the accelerated thematic search of the necessary information, including search through WWW.

The second (revised) edition of the CD-ROM "Russian Academy of Sciences. 1724-1999" was developed and published. It relies on the updated database and features also German language version. A Web site of the same name was re-developed accordingly. The site (http://hp.iitp.ru) works since July, 1998 and is being visited intensively.

A multipurpose digital model for modeling of image distortion / restoration systems has been developed. The model provides adaptation to the given configuration and parameters of the modeled system. The digital model is based on a principle of conveyor processing data flow. It has a block structure, and each block realizes a certain type of base transformations / processing of images. In each base block of the model, in turn, there are software-switchable blocks realizing a concrete kind of data transformation. All functional block can be easily re-configured at the run time. The data flow is processed consistently in the manner of block by block with pre-determined and software-changeable values of parameter. The developed model is an effective tool for experimental realization and comparison of various image restoration algorithms.

A new approach to constructing rank-order filters based on analysis of spatially connected neighborhoods is offered. Fast algorithms for adaptive constructing spatially connected neighborhoods for the proposed rank-order filters are developed. The computer experiments have shown the proposed rank-order filters outperformed the conventional rank-order filters in terms of the mean square error, the mean absolute error, and a subjective visual criterion. An extension of the approach to rank-order filtering of three-dimensional signals has been developed.

A new method of constructing the filters with improved discrimination factor for optical pattern recognition on real images is proposed. The method allows synthesizing a correlation filter with maximum diffractional efficiency, by means of zero-masking a minimum quantity of phase-only filter spectrum components. Computer simulation confirmed the superiority of the filters constructed by the proposed method comparing to the conventional phase-only filter and the optimal liner filter with respect to the discrimination capability.

Methods have been developed for expert-independent evaluation of discriminative features to automate medical image classification. Some statistics have been proposed and their threshold levels have been found to replace the subjective data of expert image analysis with the data, computed with statistics. Decision rules have been developed on the base of measured statistic values. Experiments confirmed the efficacy of proposed methods for expert-independent feature evaluation.

A new approach is proposed and new algorithms for three-dimensional visualization of data based on the imitation of the reflection from gradient surfaces have been developed The proposed approach allows visualization of real medical tomographic data by using additional resources of cognitive human recognition and diagnosis. Fast algorithms for geometric transform in 3D tomographic space has been proposed and realized. Computer experiments with real tomograms have been fulfilled.

1. Karnaukhov V.N., Wenger E., Haidinger A., Merzlyakov N.S., Zhang Y.J. An Integrated System for Digital Processing and Identification of Watermark Images // Proc. of First International Conference on Image and Graphics, August 16-18, 2000, Tianjin, China, p. 119-122.
2. Aizenberg I., Aizenberg N., Bregin T., Butakov C., Farberov E., Merzlyakov N., Milyukova O. Blur Recognition on the Neural Network based on Multi-valued Neurons // Proc. of First International Conference on Image and Graphics, August 16-18, 2000, Tianjin, China, p.127-130.
3. Karnaukhov A.V., Merzlyakov N.S., Milukova O.P. Multifunctional Digital Model of Image Blurring & Restoration Systems // Proc. of First International Conference on Image and Graphics, August 16-18, 2000, Tianjin, China, p. 147-150.
4. Gao Y.Y., Zhang Y.J., Merzlyakov N.S. Semantic-based Image Description Model and Its Implementation for Image Retrieval // Proc. of First International Conference on Image and Graphics, August 16-18, 2000, Tianjin, China, p. 657-660.
5. Karnaukhov V.N., Merzlyakov N.S., Rubanov L.I. Integration of Image Processing and Database Management Systems // Proc. of the First International Conference on Image and Graphics, August 16-18, 2000, Tianjin, China, p. 665-668.
6. Rubanov L.I., Karnaukhov V.N., Kuznetsov N.A., Merzlyakov N.S. Interactive Systems for Digital Processing, Storing and Displaying of Archival Images. Proc. of the 3rd International Conference "Digital Image Processing and its Application" (November 29 – December 1, 2000), Moscow, RNTORES – ICS RAS, 2000, p. 118-123.
7. Karnaukhov A.V., Merzlyakov N.S., Milukova O.P. Iterative Restoration of Color Images // Proc. of the 3rd International Conference "Digital Image Processing and its Application" (November 29 – Dece8mber 1, 2000), Moscow, RNTORES – ICS RAS, 2000, p. 124-128.
8. Karnaukhov V.N., Merzlyakov N.S., Milyukova O.P. Multi-functional digital model of the image blurring and restoration // Computer Optics. 2000. No. 20. P. 118-121.
9. Kober V., Choi. T.S. Single-output multichannel pattern recognition with projection preprocessing // Optical Engineering. 2000. V. 39. No. 8. P. 1-10.
10. Kober V. Speech enhancement using short-time discrete cosine transform // Pattern Recognition and Image Analysis. 2000. V. 10. No. 2. P. 1-5.
11. Kober V., Seong Y.K., Choi T.S., Ovseyevich I.A. Trade-off filters for optical pattern recognition with nonoverlapping target and scene noise // Pattern Recognition and Image Analysis. 2000. V. 10. No. 1. P. 149-151.
12. Kober V., Ovseyevich I.A. Phase-only filter with improved filter efficiency and correlation discrimination // Pattern Recognition and Image Analysis. 2000. V. 10. No. 4. P. 514-519.
13. Mozerov M., Kober V., Ovseyevich I.A., Choi T.S. Motion stereo matching using a modified dynamic programming // Pattern Recognition and Image Analysis. 2000. V. 10. No. 1. P. 90-96.
14. Kober V., Mozerov M., Alvarez-Borrego J., Ovseyevich I.A. Rank image processing using spatially adaptive neighborhoods // Pattern Recognition and Image Analysis (in print).
15. Kober V., Mozerov M., Alvarez-Borrego J., Ovseyevich I.A. Fast algorithms of rank-order filters with spatially adaptive neighborhoods// Pattern Recognition and Image Analysis (in print).
16. Mozerov M., Kober V., Choi T.S. Improved Motion Stereo Matching Based on a Modified Dynamic Programming // Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR’2000), June 18-22, 2000, Hilton Head, South Carolina, p. 501-505.
17. Kober V., Mozerov M., Alvarez-Borrego J. Adaptive image processing using rank-order filters with spatial connectivity of elements // Proc. SPIE, Annual meeting, Applications of Digital Image Processing XXIII, San-Diego, 2000 (in print).
18. Mozerov M., Kober V., Ovseyevich I.A. Real-time processing of night vision // Pattern Recognition and Image Analysis –5, Oktober 16-21, Samara, Russia, 2000, p. 651-655.
19. Belikova T.P., Stenina I.I., Yashunskaya N.I. Computer-Aided Methods to Support Image Interpretation in the Case of Uncertainty // Proc. 14 Int. Congress on Assisted Radiology and Surgery (CARS’2000). San-Francisco (USA): Excerpta Medica, Intern. Congess Series, V. 1214, p. 1050.
20. Belikova T.P., Stenina I.I., Yashunskaya N.I. Computer-aided methods to recover strategies for visual search and navigation // SPIE Congress on Medical Imaging. Medical Imaging 2000, San-Diego, USA, Proced. SPIE, V. 3981, 2000, p. 240-250.
21. Palenichka R., Belikova T., Ivasenko I. Robust extraction of diagnostic features of lesions from medical diagnostic images // Machine Graphics and Vision. 2000. V. 9. No. 1/2. P. 475-485.