基于活动轮廓模型的图像分割算法分析-计算机技术专业论文.docxVIP

安徽大学硕士学位论文Ab 安徽大学硕士学位论文 Abstract PAGE PAGE IV Abstract Image segmentation is one of the key problems in computer vision. The main task of it is to partition the image into several different parts and then transform them into more abstract and concise forms, which make high level image analysis and comprehension possible. Although people all along put great emphasis on the study of image segmentation, most traditional image segmentation methods are based on the edges derivative in local windows or simply statistical algorithms(for example histogram),which results in the limitation of themselves such as the non-integrity for the edges of segmentation region and scarce of the ability for integrating prior knowledge and so on and can not satisfy the requirement of complex segmentation applications. Since 1990s, image processing methods based on Partial Differential Equation (PDE) acquired biggish achievement and greatly promote the combination of image processing methods and advanced mathematics. PDE provides a framework which can organically combine the local and global attributes and make use of prior knowledge to optimization. Active contours driven image segmentation techniques is an important type in this framework, especially the introduction of level set, which turn the local and static differential operator into global and active differential geometric can ensure preferable effect for complicated segmentation application. In this paper we synthetically expatiate several active contours driven image segmentation techniques, such as parametric active contour model and geometrical active contour model combining level set theory and C-V model based on Mumford-Shah theory presented by Chan and Vese. Lastly, we introduce the curve evolution model based on edge flow-driven theory presented by Sumengen and so on. Among them, the last two models are mainly discussed. The character which C-V model depends on is the global information of the homogeneous regions of image