Segmentation Toolbox for Tomographic Image Data

Hildur Einarsdottir

Research output: Contribution to conferencePosterResearch

150 Downloads (Pure)


Motivation: Image acquisition has vastly improved over the past years, introducing techniques such as X-ray computed tomography (CT). CT images provide the means to probe a sample non-invasively to investigate its inner structure. Given the wide usage of this technique and massive data amounts, techniques to automatically analyze such data becomes ever more important. Most segmentation methods for large datasets, such as CT images, deal with simple thresholding techniques, where intensity values cut offs are predetermined and hard coded. For data where the intensity difference is not sufficient, and partial volume voxels occur frequently, thresholding methods do not suffice and more advanced methods are required.

Contribution: To meet these requirements a toolbox has been developed, combining well known methods within the image analysis field. The toolbox includes cluster-based methods to automatically determine parameters of the different classes present in the data, and edge weighted smoothing of the final segmentation based on Markov Random Fields (MRF). The toolbox is developed for Matlab users and requires only minimal background knowledge of Matlab.
Original languageEnglish
Publication date2014
Number of pages1
Publication statusPublished - 2014
Event3rd Annual Conference on Body and Carcass Evaluation, Meat Quality, Software and Traceability (FAIM 2014) - Danish Technological Institute, Taastrup, Denmark
Duration: 24 Sep 201426 Sep 2014
Conference number: 3


Conference3rd Annual Conference on Body and Carcass Evaluation, Meat Quality, Software and Traceability (FAIM 2014)
LocationDanish Technological Institute
Internet address

Bibliographical note

Published in 'Farm Animal Imaging' a report of the 3rd annual conference of the COST Action FA1102 (FAIM III)


Dive into the research topics of 'Segmentation Toolbox for Tomographic Image Data'. Together they form a unique fingerprint.

Cite this