Balovsyak S. Multilevel methods for processing of electron-diffraction and X-ray signals in computerized information-measuring systems.

The dissertation devoted to solving the scientific and technical problem of increasing the accuracy and computation speed of processing of electron diffraction and X-ray signals in computerized information-measuring systems (CIMS). The purpose of the work is to improve the accuracy and speed of processing the experimental signals in the CIMS by multilevel methods. The conception of a multilevel approach to the processing of experimental signals was proposed, which consists in the calculation and analysis of additional levels of signals, that provides the increase of speed or accuracy of the developed methods in order. The set of methods have been developed based on the proposed conception. The method of analysis and synthesis of intensity distribution of profiles of images was developed, the characteristic feature of which is the use of conical sections as an envelope series of profiles, which allows to increase the accuracy of the calculation of parameters for the averaged profiles in order. High-precision methods for multilevel interpolation of signals were proposed, in which interpolated signals consist of the sum of corrected and matching functions. A high-speed method of increasing the local contrast of images is developed, which uses the value of the envelopes of minimum and maximum signal values within the local windows. The method of adaptive oriented filtration of images in a spatial domain, which uses the oriented two-dimensional Gaussian distribution as the core of the filter, was proposed. A mathematical model and method of oriented filtration of band images in the frequency domain, which is based on the spatial allocation of the bands and their subsequent filtration, is proposed. A high-precision method of combining image objects with the use of a genetic algorithm was developed. The method of multilevel analysis of energy spectra of images was proposed, which provides a small error of the calculation of the average spatial frequency of the image. Get further development: the high-precision method of detecting the spatial position of straight lines, circles and ellipses on images; high-precision and multi-level methods of automatic level detection and removal of Gaussian noise on images; wavelet-filtration method using the family of biorthogonal wavelets; high-speed method for training artificial neural networks; a high-precision method of deconvolution of images, which as a point spread function uses an oriented two-dimensional Gaussian distribution. The practical importance of the obtained results is that on the basis of the obtained theoretical positions and methods the hardware-software of CIMS on the basis of X-ray diffractometers and electronic microscopes were developed. The results of the dissertation research were implemented in 3 research institutes of the National Academy of Sciences of Ukraine (Kyiv) for processing of experimental electron diffraction and X-ray signals, into the educational process of Chernivtsi National University.