The purpose of my thesis is the detection, delimitation and classification of linear structures in mammographic images. I scrutinized the topic from every point of view, I examined all the solutions given by the technical bibliography relating to this matter, and finally I chose the methods that were suitable for me to solve the task efficiently.
I made a binary edge image during solving the task, which contained the detected linear structure. The production of the edge image is made by modules organized into pipeline hierarchy. The first module does the display-work of the essence of the mammographic images, in other words it pulls the linear structures out of their environment. The display-work was made by the widely used Gabor filter. The second module does the recognition of linear structures, which is performed by an edge-detecting method, the line operator. The third module does the filtering of essential information and the making of edge image, which is made by a method designed by me. In this module the inessential linear structures, grenules, the breast periphery and the microcalcificational specks are filtered out. The last module of the pipeline defines the typical measurements based on the edge image, measurements like the frequency of linear structures, parallelism and the number of points of intersection.
For the examination and testing of the results I created a picture database of eighty pictures that contains the typical linear structures occurring in mammographic images. I examined the working of the method on the pictures of the database, then I analysed the results I got and made my conclusions.
The method was able to well detect the linear structures in different mammographic images, it delimited the detected structures and gave an adequate characterization about them.