The Research Of New Algorithms Of The Edge Extraction And Segmentation Of Medical Image

Medical image segmentation is an important part of medical image processing. Theresults of medical image segmentation provide the basis of qualitative and quantitativejudgment for medical research and clinical treatment.Based on the studies of medical image segmentation algorithms, this paperintroduced the basic principle of image segmentation, discussed the performance of thetraditional edge detection and region segmentation algorithms. For the shortcomings ofthe traditional edge detection algorithms, the threshold-based segmentation algorithmsand the region-based algorithms, this made a research on three algorithms, including anew edge extraction algorithm, two optimized threshold-based segmentation algorithmsand an optimized region segmentation algorithm. Main works are as follows:(1) A kind of edge detection and tracking algorithm based on the consistency of theedge curve was studied. Using the feature of the consistency of the edge curve of theobject, the algorithm first detect the strong edge, and then use the pre-set track strategyto track the weak edges in the image to achieve the purpose of extracting the edge.(2) Two threshold segmentation algorithms based on the two-dimension gray-scalehistogram were studied, including the2-D Otsu algorithm and2-D maximum entropyalgorithm. Both of the original Otsu algorithm and the maximum entropy algorithm arebased on on-dimension gray-scale histogram. Based on1-D gray-scale histogram, thispaper studied the2-D Otsu algorithm and the2-D maximum algorithm and thendiscussed the possibility and feasibility of the algorithm in higher dimension.(3) A watershed segmentation algorithm based on control marker was studied. Theexistence of a large number of minimum areas will lead to the phenomenon ofover-segmentation in direct watershed transform; this phenomenon will make thesegmentation meaningless. This paper studied a watershed algorithm based on thecontrol marker to overcome the shortcoming above. The algorithm uses the internal andexternal marks to eliminate the minimum area to overcome the problem ofover-segmentation.The results showed that the edge detection and tracking algorithm in this paper caneffectively avoid the complex problem of threshold selection, and has an advantage innoise immunity, the continuity and smooth of the edge and single-pixel over thetraditional edge detection algorithm; compared to1-D Otsu algorithm, the2-Dalgorithm in this paper can effectively reduce the error-divide probability; compared to the original watershed algorithm, the optimized watershed algorithm can effectively theover-segmentation problem to improve the effect of image segmentation.