| With the rapid development of computer technology, besides graphic technologyhas been widely used, both the quality and demonstrating ways of medical image havebeen greatly improved. Thus by the image processing and analysis technology, thediagnosis and treatments make enormous improvements. Medical image segmentationis a key technology of quantitative information which to extract special organizationfrom video image. Meanwhile it is also the pretreatment and precondition to realize thereconstruction of medical3D. In the article, several common image segmentationmethods have been introduced and compared separately, finally we summed up theadvantages and the opposite of these methods. We applied a segmentation algorithmwhich is based on the regional growth methods to the target images’ preprocessing, andalso used contour tracing methods and mathematical morphology to make aoptimization analysis and improvement to the segmentation’ results. Furthermore, wemade a quantitative and feasibility analysis to the contour tracing methods, and also afeasibility analysis to the image processing results of mathematical morphology. Basedon fuzzy connected graph and regional growth of image segmentation, a new methodwas put forward by calculating the attribute similarity variance method to obtain theoptimal seed points, utilizing optimal seed points to construct fuzzy graph, and using theregion growing algorithm to obtain experimental results on the basis of the fuzzy graph.The template, obtained by iterative adaptive threshold algorithm, is proposed to getrange of limiting area growth. Using this algorithm, better segmentation results can beobtained in segmented object boundary vague and target area non-uniform graydistribution circumstances.The contour tracing method is a kind of relatively simple and swift algorithm toextract images’ outline. We had worked out relevant procedures, extracted the images’outline after the segmentation by regional growth methods, however we couldn’t judgethe results which tested by the algorithm were accurate and reliable, and could not showthis algorithm’s applicable range and accuracy through quantitative points. In order tosolve this problem, we put forward an idea of fitting, from the quantitative points, to fitout the original images’ outline curve and use tracking method to get the outline curverespectively. We could make a judgment to the fitting degree of the extracted outline and its original images through two curves’ equations or their function values so that tomake an analysis that if the algorithm is accurate or reliable.Mathematical morphology is a common and strict image processing method. Wewould first make a detail introduction to the basic operators of mathematicalmorphology，including two value expansion, binary corrosion, binary open shut andgray inflation, gray corrosion, gray open and close operator, and list every operator’sproperties and summarize their operation effects. Then used the mathematicalmorphology’ basic operation to make a post-processing to the segmented images. Forthe results which existing a hole or projection, we adopted different structures to make adifferent mathematical morphological operation. In the course of experiment we foundthat different structure elements could lead to significantly distinction to the results. Theexperiment’s data showed that the structure element was a most key factor to influencethe image processing. Finally, in order to deal with the image feature, we should selectmathematical morphology operator and structure elements’ size flexibly so that toachieve the best effect. We found that it was effective and practicable to use theproposed methods to test the images’ post-processing results by mathematicalmorphology.Combined with the images’ data to work out procedures to complete the operations,we concluded that adopted the image processing methods could achieve the goal. It waseffective to segment out the bones image in the MRI to make further medical research,it has a certain theoretical significance. |
PDF Full Text Request