Vyshnevyi S. The methods of two-stage joint filtration and segmentation of heterogeneous texture images

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0413U005836

Applicant for

Specialization

  • 05.12.17 - Радіотехнічні та телевізійні системи

21-10-2013

Specialized Academic Board

Д 26.002.14

Publishing and Printing Institute of Igor Sikorsky Kyiv Polytechnic Institute

Essay

The aim of dissertation research is to design methods of causal and non-causal two-stage joint filtration and segmentation that allow to increase efficiency of filtration and segmentation of heterogeneous texture images distorted by interference with independent values of samples compared with one dimensional methods, with slight increase of computational costs. The object of research is filtration and segmentation of heterogeneous aerospace images distorted by interference with independent values of samples. The subject of research is methods of filtration and segmentation of heterogeneous texture images distorted by interference with independent values of samples. In dissertation first developed method of causal two-stage joint filtration and segmentation of heterogeneous texture images distorted by interference with independent values of samples in which at first stage one-dimensional joint filtration and segmentation is processed across rows and columns from their beginning, with consequent jointing calculated a posteriori distributions at second stage, that allows to increase efficiency of processing compared with one-dimensional methods, with slight increase of computational costs.First developed method of non-causal two-stage joint filtration and segmentation of heterogeneous texture images distorted by interference with independent values of samples, which allows to use all observations in row and column intersecting in a current point and to increase efficiency of processing compared with method of causal two-stage joint filtration and segmentation with slight increase of computational costs. Improved methods of causal two-stage filtration of homogeneous grayscale and gaussian images distorted by interference with independent values of samples, which does not need additional computation of a posteriori probabilities and moments of single-point distributions in every point of image, that allow to decrease computational costs without decreasing of filtration quality. The results of dissertation research are inculcated in work of state enterprise "Central scientific-research institute of navigation and control" (act №93 from 04 of June, 2013).Basic theoretical backgrounds of dissertation research are used in an educational process in subject "Optimization of radio system development" (act №94/2100 from 21 of May, 2013).

Files

Similar theses