Patrick Sturm

3D-color-structure-code: a hierarchical region growing method for segmentation of 3D-images

The Color Structure Code (CSC) [RE98] is a very efficient segmentation method for color images. It operates with a hierarchical region growing method. Originally, the CSC was constructed to work on two dimensional images. The generalization of the CSC for 3d images, is not a trivial task. Some important modifications of the original (2d) algorithm have to be done as the under lying 2-dimensional hierarchical hexagon topology cannot be generalized in all aspects to three dimensions.

Hierarchical cell structures for segmentation of voxel images

We compare three hierarchical structures, S15, C15, C19, that are used to steer a segmentation process in 3d voxel images. There is an important topological difference between C19 and both others that we will study. A quantitative evaluation of the quality of the three segmentation techniques based on several hundred experiments is presented.

3D Segmentierung mittels hierarchischer Inselstrukturen

In vielen Bildverarbeitungsaufgaben stellt die Segmentierung einen wichtigen Schritt zur analytischen Auswertungsphase dar. Ihr Ziel im Dreidimensionalen ist es, einen Volumendatensatz in eine Menge von Regionen aufzuteilen, die in Bezug auf gewisse Eigenschaften homogen sind. Das Ergebnis dient meist als Basis fur eine anschliesende Klassifizierung. Die hier vorgestellte Segmentierungsmethode ist ein inharent paralleles Regionenwachstumsverfahren, das in hierarchischen, einfach uberlappenden 3D Inselstrukturen arbeitet und die Vorteile von lokalen Verfahren mit denen der globalen Verfahren verbindet. Die Regionen werden durch Verknupfungs- und Trennungsoperationen auf den verschiedenen Hierarchieebenen detektiert. Das Verfahren lauft stabil und liefert sehr gute und reproduzierbare Segmentierungsergebnisse, welche mit Hilfe von modernen 3D Visualisierungstechniken prasentiert werden.

3D-Color-Structure-Code – A New Non-plainness Island Hierarchy

The Color Structure Code (CSC) [5] is a very fast and robust region growing technique for segmentation of color or gray-value images. It is based on a hierarchical hexagonal grid structure of the 2d space that fulfills several interesting topological properties. It is known that not all of these properties can be fulfilled together in 3d space. Here we introduce a new 3d hierarchical grid structure that fulfills the most interesting properties. A 3d CSC-segmentation based on this grid structure has been implemented.

Properties of a Three-Dimensional Island Hierarchy for Segmentation of 3D Images with the Color Structure Code

The CSC a very robust and fast color segmentation method. To do a real 3d segmentation of voxel images with the CSC, we have to replace the hexagonal island hierarchy by a 3d island hierarchy with the same properties. The sphere island hierarchy which is defined on the most dense sphere package can be used for segmentation with the CSC algorithm. Unfortunately, the sphere island hierarchy cannot fulfill all properties of the hexagonal island hierarchy at the same time.

A CSC Based Classification Method for CT Bone Images

The CSC (color structure code) is a robust and fast two dimensional segmentation method which has been already generalized to three dimensional images. As the CSC does not need any prior knowledge it can be used for different applications. In this paper we focus on the segmentation of bones from computer tomography data (CT) with the CSC. In the postprocessing step CSC segments will be classified according to their average hounsfield value. The classification is steered by some application specific topological rules.

Analysis of brain images using the 3D-CSC segmentation method

The 2D segmentation method CSC (Color Structure Code) for color images has recently been generalized to 3D color or grey valued images. To apply this technique for an automated analysis of 3D MR brain images a few preprocessing and postprocessing steps have been added. We present this new brain analysis technique and compare it with SPM.

3D-CSC: A General Segmentation Technique for Voxel Images with Application in Medicine

The successful 2d segmentation method CSC has recently been generalized to 3d. We shortly introduce the concept of both 2D- and 3D-CSC and present two use cases (classification of MR brain data and CT bone data) which demonstrate that analysis of segments generated by the CSC allows high quality classification of 3d data by relatively easy means.