Application Research Of The Improved Watershed Algorithm In Medical Image Segmentation

Automatic and robust medical image segmentation method is the basis and key step to further understand and analyze the image.This dissertation based on watershed algorithm proposes a method to get the continuous boundary of the object region and it can realize the application of the watershed algorithm in medical images.The dissertation makes a survey of the main methods and techniques in the filed of image segmentation,and summarizes the respective application spheres,advantages and disadvantages. And then,dissertation makes a in-depth study of the applications of mathematics morphology and watershed algorithm based on mathematics morphology in medical image segmentation.On this basis, an improved watershed algorithm is proposed based on K-means clustering algorithm to satisfy the need of construct the numerical calculation model of the bio-electromagnetic field.First,the morphological filtering technique is applied for image pre-processing to get the morphological gradient image.And then the improved K-means clustering algorithm is used to get the multi-threshold segmentation result of the gradient image. The over-segmentation has been reduced and the object region has emerged,then the methods of scan line seed fill algorithm and mathematics morphology are used to to get the object template.Afterward,the edge information of original image gained by Canny operator combined with mathematical morphology operators is used optimizing initial segmentation template for achieving the segmentation result which is more conform to the target demand.Last,using the region relevancy of neighboring frame,the accurate segmentation result of the preceding image is taken as the object region reference template of the neighboring frame.The edge information of original image gained by Canny operator and mathematical morphology operators are combined to correct the object region reference template,and then to get the final segmentation results.The continual segmentation of image sequence is realized,thus the computation is reduced,the segmentation efficiency is enhanced.The proposed method is applied to the MRI of the thorax and brain, experimental results show that this proposed method has an impressive effect on over-segmentation,and reduces the manual participation,enhances automatic of algorithm.