基于分水岭算法的磁共振脑图像自动分割

Watershed-Based Brain Magnetic Resonance Image Automated Segmentation

基于分水岭算法,提出了一种新的非脑组织去除和自动的脑磁共振图像的分割方法,利用区域合并技术克服分水岭算法固有的过分割问题.通过参数的设置,可以将图像中的非脑组织去除掉;对已去除非脑组织的图像,巧妙地将分水岭算法、区域合并和k-均值算法相结合,可进行全自动地分割,效果良好.

Based on watershed algorithm a new method to remove nonbrain tissue and segment brain magnetic resonance (MR) images automatically was presented. Region merging technique was used to overcome the oversegmentation. Through appropriate choice of parameters, brain tissue can be extracted. Subsequently, a fully automated procedure was designed to segment the MR images with only brain tissue that is a combination of three existing techniques: watershed algorithm, region merging technique and kmeans algorithm from pattern recognition. Results were presented to show its performance.