Efficient filtering algorithms for extraction of patterns in images

OData support
Supervisor:
Dr. Horváth Gábor
Department of Measurement and Information Systems

The overlapped vessel opacities and the vessel branching points can often cause similar shape as the nodule opacities on the chest radiograph. For the time being, detection of the lung vessels on chest x-ray remains an open problem, therefore very few publication have been released in the literature of computer analysis of chest radiographs, although the result of such procedures could be used to reduce false candidates.

The thesis aims to implement image processing algorithm, which can be emphasize anatomical regions such as vessel opacities or their branching points in chest radiographs effectively. In my thesis, I deal with the implementation and evaluation of vessel and junction filters, which can be enhance the curvilinear structures on lung images.

To emphasize the vessel opacities on chest radiographs, I use a image filter which is sample the vicinity of the filter origin based on a pre-determined structuring element and it is calculate the output value from the sampled image intensities.

To enhance the quality of the lung images, I implemented pre-processing image filters and reconstruction algorithms for the digital lung tomosynthesis.

To detect the vessel opacities on lung images, I implemented image filter algorithms which are based on the eigenvalue analysis of the Hessian and the correlation matrix. The evaluation of the implemented shape filter on a benchmark fundus image database results 0.82 precision.

Downloads

Please sign in to download the files of this thesis.