Color morphological reconstruction as a segmentation tool for microscope cell images

Abstract

There are numerous segmentation methods for gray level microscope cell images, however in some situations the texture features and roughness are not as relevant as the color for the segmentation task. For example, for detection of histological analysis, the relative sizes of nucleus and cytoplasm, as well as their shapes, are the relevant features, while other characteristics such as texture and roughness have no value in the diagnosis. In this context, geodesic reconstruction is one of the image operators, of mathematical morphology, that facilitates image segmentation of individual objects. This operator uses markers to highlight, in the image, objects of interest, in order to separate them from the rest of the scene. This paper presents a new segmentation method, for microscope cell images, based on mathematical morphology color reconstruction, where the markers can be obtained automatically or semi-automatically. Automatically, when you want to detect those cells that were not removed at the generation of the markers. Semiautomatically when the expert manually selects one pixel for each cell of interest.