Image segmentation is the process of partitioning an input image into non-overlapping/disjoint regions. Various methods have been used for image segmentation. Among these methods, watershed-based algorithms have been widely utilized due to their fast computational speed. However, their sensitivity to noise, and also over-segmentation, has made watershed approaches unsuitable for noisy images. In this paper, a novel method for MRI segmentation is proposed. For this purposed, a simple and fast feature extraction method is used. Then, the feature space, is clustered by Self-Organizing Map Neural Networks (SOMNN). After that, an edge map is set up from the clustering result. Finally, watershed transformation is utilized on the edge map. Our algorithm is robust against noise. Although watershed transformation is used in our approach, a region merging and denoising algorithms are not utilized as pre- and post- processing, respectively. This significantly improves the segmentation speed.
|Number of pages||6|
|Publication status||Published - 2009|
|Event||IEEE International Conference on Signal and Image Processing Applications (ICSIPA) - Kuala Lumpur, Malaysia|
Duration: 18 Nov 2009 → 19 Nov 2009
|Conference||IEEE International Conference on Signal and Image Processing Applications (ICSIPA)|
|Period||18/11/09 → 19/11/09|
Emambakhsh, M., & Sedaaghi, M. H. (2009). Automatic MRI brain segmentation using local features, self-organizing maps, and watershed. 123-128. Paper presented at IEEE International Conference on Signal and Image Processing Applications (ICSIPA), Kuala Lumpur, Malaysia. https://doi.org/10.1109/ICSIPA.2009.5478631