AUTOMATIC MESH SEGMENTATION USING ATLAS PROJECTION AND THIN PLATE SPLINE Application for a Segmentation of Skull Ossicles Makram Mestiri, Sami Bourouis and Kamel Hamrouni National Engineering School of Tunis, University of Tunis, El Manar, Tunis, Tunisia Keywords: Mesh, 3D segmentation, Registration, Thin-plate spline, Atlas, Skull bones. Abstract: Mesh segmentation has become a crucial step in many computer graphics applications. This paper provides new method for three dimension Atlas based mesh segmentation using thin plate spline approach (TPS) and a new FNN algorithm. This method consists of three steps: first, we apply a rigid registration between two meshes the atlas and the mesh to segment. The second step is the application of an elastic registration using thin plate spline method. The last step is the identification of the different regions to segment the mesh using our FNN algorithm. We tested the performance of our method on synthetic images and on a real human skull and found that the preliminary results obtained are satisfactory. 1 INTRODUCTION In medical imaging, the use of three- dimension shape has greatly facilitated the disease diagnosis. Segmentation is an important and difficult step in medical images analysis. The segmentation of polygonal meshes can divide the mesh into multiple segments in order to simplify or change the mesh representation to another representation more meaningful and easier to analyze. There was a wealth of research focused on methods for polygonal meshes segmentation we can classify them into two categories: methods based on geometric features, and methods based on semantic approaches. In the first case, the shape is segmented into a number of uniform patches with respect to some surface properties, while in the second one the segmentation is aimed at identifying relevant features of the shape. In patch segmentation methods, shapes are dividedinto regions that have certain geometric features such as flatness, convexity, approximation to a primitive bend (Shamir, 2008). Among the most frequently used algorithms, we cite the growth of regions, watershed, deformable models, hierarchical partitioning, spectral partitioning and skeletonization. Segmentation into meaningful parts: This type of segmentation can divide the object into meaningful components. This segmentation is mainly based on human perception. Thus, some researchers propose to use primitive specifying the 3D shape such as boundaries to break down a scene or object (Lon, 2007).Others use models (Atlas) to project the predefine segmentation of the atlas to the object to be segmented (Commowick, 2010). The atlas based segmentation has become a standard method for brain segmentation Oliver (Oliver, 1998) summar- izes the segmentation of the brain atlas in three stages. The first step is to match the image overall patient and image atlas for which they are located in the same repository. The second step is to apply a local deformation to bring the two images perfectly. Finally a transformation will be applied to the atlas image. An illustration showing figure1 the difference between the two method (surface patches and significant parts). a b Figure 1: (a) Segmentation using patch approaches (Shamir, 2008), (b) Segmentation into meaningful parts (Katz, 2006). 221 Mestiri M., Bourouis S. and Hamrouni K.. AUTOMATIC MESH SEGMENTATION USING ATLAS PROJECTION AND THIN PLATE SPLINE - Application for a Segmentation of Skull Ossicles. DOI: 10.5220/0003371702210225 In Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP-2011), pages 221-225 ISBN: 978-989-8425-47-8 Copyright c  2011 SCITEPRESS (Science and Technology Publications, Lda.)