Multi-Level Single Phase Level Set Method In Medical Image Segmentation

The single phase level set method is a very effective and easy-to-implement method in image segmentation. However, due to the disadvantage that it can only segment the image into two parts, its applications are limited. For images with detailed structures, the single phase level set method cannot extract subtle structures satisfactorily. To overcome this disadvantage, we proposed a multi-level single phase level set method. This method has several levels of segmentation, after each one, the result sub-area is segmented again. While keeping irrelevant areas away, the segmentation result will be refined and the exact fine detail of the image can be extracted. Meanwhile, its implementation is convenient. With the method developed in this article, the initial level set function for the fine levels has little restriction. For grey scale images constituted of different areas with grey level in normal distribution, multi-level single phase level set method performs very well. At last, we carry out some analysis of different image features commonly used in image segmentation. For chest CT images investigated in this paper, these features have a good presentational ability for different areas. For the chest X-ray images, however, since the image of different organs lap over each other and the image structure is complicated, the performances of these features are all poor , they still need improvement.