Compound Vector Based Improved ACM Model And The Research Of Medical Image Segmentation

The main purpose of the image segmentation is to extracting the region ofinterest(ROI) from the background, or partitioning the different ROIs intononoverlapping regions. Image segmentation is the key step to realize the researchfrom general image processing into image analysis. It is very important to imagepreprocessing and pattern recognition, such as feature quantification, imageregistration 3D reconstruction and etc. so it has been an important and challengingproblem for many decades.At present, the active contour model has become an important tool of themedical image analysis. It incorporates the data constraints derived from imagesand the prior knowledge about the ROIs into the variational framework and itsapplication has covered image denoising, image segmentation, image registration,image repairing, surface reconstruction and motion tracking, etc. First of all, we summarizes the actual state of image segmentation. In this part ,we provides anoverview of active image segmentation methods. Chapter 3 introduces the idea ofactive contour model, and the evaluation of traditional active model, especially themathematics of gradient vector flow model. The main emphasis in on Chapter 4.we performs the active model algorithm firstly, then analysis the problem by usingthe active contour model. When the classical active contour model is applied tosegment the image, an initial contour must be set near the boundary of ROI and themodel cannot segment deeply concave regions accurately. On the basis ofanalyzing the active contour model, a improved ACM model algorithm for imagesegmentation based on the compound vector is proposed. With the purpose ofacquiring a more ideal edge map, the image is processed by the generalized fuzzytheory, which avoids the weak edge and sharp cape part of the image blurred in thesmooth process. The model is designed by replacing the traditional vector fieldwith the compound vector .and the constraint parameters of the GVF field is madeadaptable to the target object features. The algorithm has been proved efficientthrough many of the experiments.In the end of the paper, two problems associated with medical imagesegmentation are discussed, which are the problem of objective evaluation of thesegmentation results and the problem of interaction in the segmentation. Difficultiesharassing the evaluation method are listed and the corresponding solutions arecollected. In many medical image related applications, human-machine interaction isthe sine qua non for accurate segmentation. This paper also presents a summary forthis issue and we can design a more effective segmentation algorithm under thisguidance.