Description: Increase of the amount of intelligence processing results in high time costs. Therefore, it becomes very important to auto-mate the process of selecting objects in images. That will increase the efficiency of solving intelligence problems. The main tasks that in general significantly affect the quality of the decryption process are the tasks of detection and identification of single and complex intelligence objects in the image. The main problem when converting a bitmap to a vector format is to achieve 100% similarity between the original and final images. The criteria for segmented image quality have been defined. They are the ho-mogeneity and dissimilarity of neighboring regions; the smoothness of their borders, the small number of small “holes” within the region, the boundaries of each segment should be simple, spatially accurate. The possibility of using the existing methods of constructing a segment map has been analyzed. At it the low tone saturation of space and aerial images and the presence of a large number of textured objects were taken into account. Since the range of possible color values of pixels in the textured region varies widely, the usage of decryption features based on the color characteristics of the analyzed area does not ensure accept-able decryption results. Decrement of the range of allowable values eliminates most of the errors of the first kind, but propor-tionally increases the number of second kind errors. When using the texture features of an area as a homogeneity criterion, the exact similarity of the vector image of the object is not achieved, and the closeness of the texture kernel sizes of different objects leads to first and second kind errors. The article suggests usage of the constructed field of fractal dimensions in order to deter-mine the boundaries of different segments, in which there is an impulse change of fractal dimension values.
Keywords: image segmentation, vectorization, color space clustering, textural features, contour filters, fractal dimension field.
1.Khudov, V.H., Kuchuk, H.A., Makoveychuk, O.M. and Kryzhna, A.V. (2016), “Analiz vidomikh metodiv sehmentuvannyazberezhenyy, chto otrimani z bortovykh system optyko-elektronnoho sposterezhennya” [Analysis of known image segmentation techniques obtained from on-board optical-electronic surveillance systems], Information Processing Systems, No. 9, pp. 77-80.
2.Shytova, O.V., Pukhlyak, A.N. and Drob, E.M. (2014), “Analiz metodiv sehmentatsiyi tekstur oblastey zobrazhenʹ v syste-makh obrobky zobrazhen” [Analysis of segmentation methods for texture image areas in image processing systems], Scientific Statements of BelGU, No. 8(179), pp. 182-188.
3.Jan, F., Usman, I., Khan, S.A. and Malik, S.A. (2013), Iris localization based on the Hough transform, a radial-gradient op-erator, and the gray-level intensity, Optik, No. 124(23), pp. 5976-5985, available at: www.academia.edu/5065876/ Iris_Localization_using_Circular_Hough_Transform_and_Horizontal_Projection_Folding.
4.Zhang, Chao, Changming, Sun and Ran, Su (2015), Clustered nuclei splitting via curvature information and gray-scale dis-tance transform, Journal Microscopy, No. 259(1), pp. 36-52, available at: www.ru.scribd.com/document/362374559/2015-ClusterNucleiViaCurvatureInformation-3-pdf .
5.Hu, S., Xu, C., Guan, W., Tang, Y. and Liu, Y. (2014), Texture feature extraction based on wavelet transform and gray-levelco-occurrence matrices applied to osteosarcoma diagnosis, Bio-Medical Materials and Engineering, No. 24(1), pp. 129-143, avail-able at: www.ncbi.nlm.nih.gov/pubmed/24211892.
6.Nixon, M.S. and Aguado, A.S. (2002), Feature Extraction and Image Processing, A division of Reed Educational and Pro-fessional Publishing Ltd, Oxford, 330 p.
7. Burge, M. and Kropatsch, W. (1999), A minimal line property preserving representation of line images, Computing,No. 62(4), pp. 355-368, available at: https://link.springer.com/article/10.1007/s006070050029.
8.Zenzo Di, S., Cinque, L. and Levialdi, S. (1996), Run-based algorithms for binary image analysis and processing, IEEETransactions on Pattern Analysis and Machine Intelligence, No. 1(18), pp. 83-89, available at: www.kissly.net/book/3948154204024158B8FD?utm_source=ps84&utm_medium= ifinisepykyl.tk& utm_campaign =fnom&x=198084.
9.Linda, G. Shapiro (1996), Connected Component Labeling and Adjacency Graph Construction, Elsevier, No. 1(31), 293 p.,available at: www.index-of.co.uk/Artificial-Intelligence/Computer%20Vision%20-%20Linda%20Shapiro.pdf.
10.Jung-Me, P., Looney, G.C. and Hui-Chuan, C. (2000), Fast Connected Component Labeling Algorithm Using a Divide andConquer Technique, CATA 2000 Conference on Computers and Their Applications, pp. 373-376, available at: www.pdfs.semanticscholar.org/08ac/ 52b31f3fe929c2dbb191558375630ce06f01.pdf.
11.Tarshyn, V.A., Sotnykov, A.M. and Pashchenko, R.E. (2014), “Metod operatyvnoyi pidhotovky etaloniv na osnovi frak-talʹnoyi obrobky zobrazhenʹ z vysokoyu ob'yektovoyi nasychenistyu” [The method of operational preparation of standards based on fractal image processing with high object saturation], Technical Vision, No. 1(5), pp. 2-8, available at: www.magazine.technicalvision.ru/archive/issue-1-5.
12.Pashchenko, R.E. (2006), “Fraktalʹnyy analiz protsesiv, struktur i syhnaliv” [Fractal analysis of processes, structures andsignals], Ekoperspektiva, Kharkiv, 348 p.