Sergey V. Ablameyko, PhD, DSc, SMIEEE, FIAPR, FIEE, FBEA, CEng,
United Institute of Problems of Informatics (UIPI)
National Academy of Sciences of Belarus
Surganov Str. 6, Minsk 220012
Republic of Belarus
Phone: +375 172 84 2171
DVP term expires December 2013
Professor Sergey Ablameyko (PhD, DSc, Prof, FBEA, FIAPR, FIEE, CEng, SMIEEE) in the General Director of the United Institute of Informatics Problems of the National Academy of Sciences of Belarus and Head of Image Processing and Recognition Research Laboratory at the Institute. In addition he is part-time Professor in Computer Sciences at the Belarusian State University, Minsk, Republic of Belarus.
The list of his publications includes over 300 items including 9 books in the area of Image Processing and Pattern Recognition. He was an invited and visiting staff with the universities and research centre in Italy, Japan, Sweden, Finland, England, Germany, UK, Greece and Spain. He has been on the editorial boards of such international journals as Pattern Recognition, Pattern Recognition Letters, Machine Graphics and Vision. He is also an Editor-in-Chief of 2 national research journals in Belarus.
Professor Ablameyko is Fellow of IET, Senior Member of IEEE, Fellow of IAPR as well as Fellow of two National Engineering Academies. He is the First Vice-President of the International Association for Pattern Recognition (www.iapr.org) and the President of the Belarusian Association for Image Analysis and Recognition. He was a Principal Investigator of numerous national and international projects including those sponsored by EU, NATO, and CIS funding bodies.
Interpreting Images of Line Drawings
While development continues, multimedia tools for planning and recording the results of work on complex engineering products and projects are now widely available. These tools can significantly improve communication within project teams but suffer from an input bottleneck; the information they require must be provided by hand. Although computers have now been commonplace in design, manufacture and maintenance processes for a long period of time, many projects have yet to be represented electronically. The majority of the necessary 3D and other product/design information is readily available, but in the form of paper documents, and particularly drawings. Manual input of drawings into CAD, GIS and other systems is a possibility, albeit a slow and expensive one. Scanners allow drawings to be transformed quickly and cheaply into digital images but highly developed software is required to convert those images into the formats required by multimedia systems.
The proposed presentation will concentrate on techniques for the interpretation of images of line drawings, covering the processes involved in and issues raised by the extraction of intermediate level graphical entities from engineering drawings, maps and plans.
From Cell Image Segmentation to Cancer Diagnosis
In differential diagnosis of cancer tumors, the role of generally accepted criteria of malignancy (which is based on the whole complex of quantitative indices of cell abnormality) is limited. This leads to a low efficiency of cancer detection at early stages of the disease and thus necessitates development and adoption of new more effective methods of cancer diagnosis. One of the most promising approaches to solve the problem is the transformation of qualitative indices of pathological changes in cells to a quantitative form with the help of the computer morphometry method. A very significant attention is devoted to this problem in the literature. It was shown that the use of thyrocyte nuclear area values allows differentiating adenoma from cancer.
The first morphometry methods were based on manual techniques that were not very effective. The discussed in the talk automated cytological image analysis techniques offers significantly higher performance. The automated procedures are able to analyze cytologic images, extract all features about cells, nucleus, and aggregate and compute parameters to carry out/verify diagnosis.
There are four main steps in the proposed procedure: a) image segmentation, b) vectorisation and morphometric database building, c) karyometric paremeters selection and expert system building, and finally d) cancer diagnosis. There were numerous attempts in the past to develop algorithms for segmentation of biomedical images. They were based on a wide range of mathematical tools namely: mathematical morphology, snakes, Fourier and Hough transforms, artificial neural networks, etc. However, due to the very complex nature of cytologic images, it is not always possible to select or develop methods for automatic segmentation that could be applied to extract objects correctly.
The proposed talk will concentrate on cytologic diagnosis of cancer with the help of morphometry method. The talk will also introduce a special computeraided image analyzer that includes an automated processing and binarization of color images; automatic raster-to-vector transformation and formation of biological objects; morphometric assessment of biological objects by quantitative parameters characterizing the changes of cell nuclei as well as building of expert system with aimed at cancer diagnosis.
Interpretation of Remote Sensing and Map Images
Interpretation of remote sensing images remains a difficult task and is mainly performed by using interactive techniques. To make the process fully automatic the auxiliary information obtained from digital maps have to be used. Such information can be obtained as a result map digitizing, which is based on a complex techniques involving automatic and interactive processes. The aim of this talk is to introduce main techniques currently employed in the world for map digitizing, matching digital maps with remote sensing images and object detection in these images. Special attention will be paid to vectorization and interpretation of large-size map-drawing images that include objects with different font, scale, orientation, etc. Analysis and classification of the existed approaches for map interpretation will be presented.
The following main topics will be covered:
- fast vectorization of large-size map-drawing images (as well as noise filtering, thinning, contouring, transformation into vector form, feature extraction);
- vector representation of map-drawings;
- automatic recognition of basic cartographical objects (lines, symbols, areas) including recognition of different line types, characters, textures;
- knowledge use for map recognition including knowledge types, representation and use;
- interactive map interpretation and correction of recognition errors;
- data structures for map representation;
- matching of remote sensing images and digital maps;
- detection of objects in remote sensing images by using the digitized information.