The Research On Geometric Active Contour Model And Its Applications In Image Segmentation

Image segmentation plays an important part in the medical image processing. It is also a hot topic for scholars at home and abroad. Geometric active contour model is widly used in the non-rigid object segmentation in recent years,which shows great advantage over the traditional segmentation algorithm. But the geometric active contour model has some disadvantages such as the high complexity of the algorithm and its poor convergence to the weak edges of medical images.The research of the thesis including three parts is described as follows:The theory of the traditional geometric active contour model is presented first. It mainly includes the curve evolution theory, the level set method and the fast implementation algorithm of the level set method. Meanwhile, the advantages and disadvantages of the traditional algorithm are also analized.From the recent research, it can be found that it is very hard to get satisfactory segmentation results by using only one segmentation algorithm. Therefore two new segmentation algorithms which are implemented by combining the geometric active contour model with the traditional segmentation algorithm are proposed in the thesis.Due to the poor convergence of the traditional algorithm to the edges, a more strong constraint derived from the presegmentation using meanshift method is used to the level set function to modify the level set method without reinitialization. The algorithm can obtain more accurate segmentation from the experimental results.Another segmenation algorithm which combines watershed transform and level set method is also proposed in the thesis. The new algorithm mainly depends on the region information which makes the algorithm more robust to the noise. The region level set function is defined according to the presegmentation results. Then the processing time of the algorithm does not depend on the size of the image but the presegmentation region number. The experimental results show that the algorithm can obtain satisfactory segmenation results of the medical image and it solves the problem of high complexity in processing large images to some extent.