Automatic marker determination algorithm for watershed segmentation using clustering
Mariela Azul González, Virginia L. Ballarin · Latin American Applied Research - An international journal · 2009
Biomedical image processing is a diffi- cult task because of the presence of noise, textured regions, low contrast and high spatial resolution. The objects to be segmented show a great variability in shape, size and intensity whose inaccurate segmenta- tion conditions the ulterior quantification and pa- rameter measurement. The partition of an image in regions that allow the experienced observant to ob- tain the necessary information can be done using a Mathematical Morphology tool called the Watershed Transform (WT). This transform is able to distin- guish extremely complex objects and is easily adapt- able to various kinds of images. The success of the WT depends essentially on the existence of unequivocal markers for each of the ob- jects of interest. The standard methods of marker detection are highly specific, they have a high com- putational cost and they determine markers in an effective but not automatic way when processing highly textured images. This paper proposes the use of clustering techniques for the automatic detection of markers that allows the application of the WT to biomedical images. The results allow us to conclude that the method proposed is an effective tool for the application of the WT.