Analysis of global and local intensity distributions for the segmentation of computed tomography images

Abstract

The segmentation of Computed Tomography images is an extremely challenging task due to the heterogeneities of the regions, their textured intensities, the presence of noise and the blurred edges which define the limits of the different areas. In this work, we propose a method for obtaining a segmentation of the abdominal region in Computed Tomography images based on the analysis of the global distribution of the intensity values in the whole data set and the local distribution within the neighborhood of each pixel. The global information is used to obtain an initial approximation, whereas the local data are analyzed to extract a much more accurate and complete segmentation. The combination of both scales allows obtaining a quite satisfactory segmentation in an automatic way.