A FAST LEVEL SET METHOD BASED ON DEFORMABLE MODEL FOR IMAGE SEGMENTATION
Fugen Zhou · Chinese Journal of Stereology and Image Analysis · 2003
A novel image segmentation technique based on level set approach for recovering shapes from two and three dimension image data is described. This model-based method has some attractive features comparing to the existing methods and overcomes some of theirs limitations. A fast method of Level Set called Narrow Band algorithm is realized in numerical domain. An initial closed curve is first placed inside or outside the shape to be isolated, Image is then processed by Gaussian filter, the contour of this shape is extracted automatically. Our techniques can be applied to model arbitrarily complex shapes very well, which include shapes with significant protrusions or transformable topology structure, but no a priori assumption about the object抯 topology is needed. The efficiency of the scheme is demonstrated through numerical experiments on some types of medical image, including CT/MRI and microscopy images. The experimental results also show that some parameters of this algorithm have biggish impacts on contour extracted, such as precision and smoothness.