Adaptivna segmentacija 3D volumnov

ČOTAR, BOŠTJAN (author), Divjak, Saša (mentor), Marolt, Matija (co-mentor)

Cilj diplomske naloge je razviti metodo za segmentacijo 3D modela drevesne strukture žil za program NeckVeins. S primerjavo meje praga in vrednosti vsakega posameznega oglišča voksla, iz 3D matrike določimo, kateri voksli predstavljajo telo drevesne strukture žil, kateri predstavljajo ozadje in kateri predstavljajo tako imenovane »sosede«. Za bolj natančen izris 3D modela, je v metodi implementirano rekurzivno deljenje sosednjih vokslov na osem manjših segmentov. Za vsakega od manjših segmentov se ponovno določi, v katero kategorijo spada. Na ta način dobimo model drevesa žile, zgrajenega iz vokslov različnih velikosti. V diplomskem delu je tako na novo razviti algoritem implementiran v delovanje programa NeckVeins, ki že vsebuje svojo metodo za segmentacijo. Na koncu je prikazana primerjava vseh algoritmov za segmentacijo v programu NeckVeins. Za primerjavo razlik med algoritmi se uporabljajo zaslonske slike 3D modelov ožilja. Primerjamo zgradbo žil, stopničavost in natančnost izrisa.

• Language: Slovenian
• Keywords: segmentacija, 3D model, Marching Cubes, voksel
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2014
• PID: 20.500.12556/RUL-29535
• Publication date in RUL: 22.09.2014

Secondary language

• Language: English
• Title: Adaptive segmentation of 3D volumes

Abstract

The main objective of the thesis was to develop a method for segmentation of a 3D vascular tree model for the program NeckVeins. By comparing the threshold value with the value of each vertex of the segment, which is a voxel, the method determines which voxels represent the main body of the vein tree, which represent the background and which represent the so-called »neighbours«. Depending on how detailed we want to plot the 3D model, the method recursively divides the adjacent voxels to eight smaller voxels. For each of the smaller segment we re-determine to which category it belongs, so that in the end the vein tree model is built from voxels of different sizes. The newly developed algorithm is implemented in the program NeckVeins, which itself already contains its own segmentation method. We compare the differences between the algorithms using screenshots of the 3D vascular model, paying attention to the structure of the blood vessel, staircasing and the accuracy of the plot.

• Keywords: segmentation, 3D model, Marching Cubes, voxel