Evaluation of Morphological Reconstruction, Fast Marching and a Novel Hybrid Segmentation Method.

An evaluation of two traditional segmentation algorithms of Morphological Reconstruction and the Fast Marching method along with a novel hybrid segmentation approach is presented. After introducing the Fast Marching and the Morphological Reconstruction segmentation, we propose a novel hybrid segmentation approach in multi-stage, which is derived from both an improved Fast Marching method and the Morphological Reconstruction. To demonstrate the effectiveness and accuracy of the three methods, we employ an MRI brain image in our experiments, in which "gold standard" is known. The evaluation is measured accordingly in accuracy and speed when running a 2.0 GHz based windows XP PC. The accuracy results of average 0.9738, 0.6302 and 0.9734 measured in similarity indexes of the Morphological Reconstruction, the Fast Marching and the hybrid approach are achieved, respectively. The computing performance required 188.6, 22.3 and 43.4 in seconds accordingly. [ABSTRACT FROM AUTHOR]