Matematický model pro automatizovanou analýzu RTG pozadí

Abstract

This bachelor thesis describes segmentation of X-ray background. Theoretic part includes fundamental knowledge about X-ray, bone imaging via X-ray and research of algorithms for X-ray image segmentation. Practical part involves selection of appropriate algorithms – OTSU, K-means and Fuzzy thresholding, designing the algorithm and it´s final implementation for X-ray background segmentation and pixel intensity extraction in MATLAB. Three specific values were tracked – mean, mode and median. For result evaluation there were created representative regions of interest from X-ray background, these regions are suppossed to replace hardly obtainable golden standard. Verification of results was carried out through calculation of absolute error of pixel intensity between global segmentation algorithm and representative regions of interest from X-ray background, which were taken from the same image. The most reliable algorithm for X-ray background segmentation was Fuzzy thresholding, due to it´s lowest absolute error of tracked values. The differences were minor, in order of tenths of intensity level which is not noticeable by human eye. It can be said, that all tested algorithms perform the segmentation of X-ray background reliably.