Browsing by Subject "skeletonization"

Now showing 1 - 4 of 4

Access status: Open Access ,
3D skeletonization of pulmonary airway tree structures
(Wydawnictwa AGH, 2010) Postolski, Michał; Janaszewski, Marcin Sławomir; Jopek, Łukasz; Babout, Laurent
Quantitative analysis of the human airway trees is a challenge in image processing and analysis. Results obtained in previous work on this subject are still not sufficient and they need improvements. Quantitative description of an airway tree consists in application of several steps: segmentation of the tree, skeletonization, decomposition and anatomical labelling, cross section generation and finally quantitative measurements. Each step needs to use different kinds of image processing algorithms. The most works were performed on different segmentation strategies, however skeletonization algorithms are still not enough tested and they require attention. Moreover, skeleton has very large impact on the quality of quantitative measurement of a bronchial tree. In the paper authors tested three skeletonization methods which are based on thinning - the most popular skeletonization approach. However, each algorithm uses different thinning strategy. The paper presents basic concepts of the algorithms and discussion about their primary features based on acquired results from real human CT images.
Access status: Open Access ,
Automatyczny pomiar długości korzeni siewek pszenicy z hodowli hydroponicznej przy wykorzystaniu metod przetwarzania i analizy obrazów
(Wydawnictwa AGH, 2009) Gocławski, Jarosław; Sekulska-Nalewajko, Joanna; Gajewska, Ewa Grażyna; Wielanek, Marzena Katarzyna
In the paper an automatic method of length measurement has been presented for the roots of wheat from hydroponic culture, based on image processing and analysis. Evaluation of this feature of root systems is important for the estimation of plant tolerance to environmental factors influencing their productivity. In the proposed method simplified procedure of root preparation is applied and low-cost equipment is assumed consisting of typical scanner device and personal computer. The method includes root image segmentation in colour space, binary image skeletonization and then 3D trees topology reconstruction from 2D skeletons of all individual fibrous roots defined as 1. order roots. The analysis of tree data structures enables the reconstruction of axes of 1. order roots and their lateral branches in the image raster. Root lengths are calculated along these axes following cubic splines smoothing of their routes. The measurements of 1. order roots, carried out for the series of 10 images (about 50 roots) revealed only small differences from the results obtained with manual method - 1,94% in average.
Access status: Restricted ,
Badanie wpływu szumów obecnych w obrazie na dokładność detekcji krawędzi za pomocą filtru wariancyjnego
(Data obrony: 2019-01-25) Pazdalski, Łukasz
Wydział Geologii, Geofizyki i Ochrony Środowiska
Celem pracy była analiza dokładności detekcji krawędzi przy pomocy metody wykorzystującej filtr wariancyjny. Jako dane wejściowe wykorzystano zdjęcia rentgenowskie przedstawiające kręgosłup. Implementacja metody została napisana w środowisku Matlab. Badania przeprowadzono dla trzech obrazów, różniących się poziomem zaszumienia oraz detali. Wyniki porównano z metodą wykrywania krawędzi za pomocą operatora Sobela.
Access status: Open Access ,
Zamykanie otworów w trójwymiarowych obiektach wolumetrycznych
(Wydawnictwa AGH, 2009) Janaszewski, Marcin Sławomir; Babout, Laurent; Postolski, Michał; Jopek, Łukasz
Hole closing in 3D volumetric objects is a challenge in computer science because from a topologieal point of view a hole is not a subset of 3D space. Therefore it is impossible to close a hole by the use of classical methods based on connected component labeling or suppressing filters. The article presents the new algorithm of hole closing based on the algorithm presented by Aktouf e.g. in 2002 year. Algorithm presented in the paper has several advantages in comparison with the Aktouf's approach: only holes are closed but not cavities - bounded by an object disconnected fragments of background. Moreover generated patch which closes a hole corresponds to its 'geometry' and geometry of the patch is not influenced by branches of the object which are situated close to the hole. The algorithm has been tested on artificially generated objects and very complicated 3D objects which represent stress-corrosion crack to be propagated in stainless steel. The article includes also results of the algorithm tests, discussion of its properties and possible applications.