_基于区域合并的纹理图像分割—MSRM算法.doc

毕业设计 题目： 基于区域合并的纹理图像分割—MSRM算法的MATLAB实现  摘 要 图像分割是图像分析及计算机视觉系统中的重要环节，是图像处理研究中的一个基本难题。图像分割是由图像处理到图像分析的关键步骤，只有在图像分割的基础上才能对目标进行特征提取、参数测量和识别，使得更高层的图像分析和理解成为可能，图像分割质量的好坏直接影响后续图像处理的效果。因此，可以说图像分割是图像处理中最为重要的环节。 基于区域的图像分割方法，将图像按内容划分成许多区域。虽然存在过分割，但是可以通过研究改进算法减少过分割或选择有效的后处理算法得到有用的结果。例如，在Mean Shift和Watershed这两种图像分割算法中，一方面可以研究各种减少过分割的改进算法。另一方面，也可以采用有效的预处理，去除噪音，使图像适宜于Watershed或Mean Shift算法分割。 MSRM是基于最大相似度的区域合并算法,该算法简单有效,不需要设定区域合并的阈值,且对单目标和多目标图像都能正确分割。本设计在提取了图像的颜色特征和纹理特征之后，计算相邻区域的相似度，在人工交互信息的指导下，基于最大相似度准则逐步对初始过分割区域进行合并，分离出图像中的目标和背景。 关键词： MSRM；区域合并； 交互式图像分割；算法；纹理图像  Abstract Image segmentation is the important elements of image analysis and computer vision systems ，and it is a fundamentalproblem in image processing。 Image segmentation is a key step that from image processing to the image analysis。Only on the basis of image segmentation to object feature extraction, parameter measurement and recognition，which make it possible to analysis and understanding of the higher-level image, and the quality of image segmentation directly affects the effect ofsubsequent image processing. Therefore, we conclude that image segmentation is the most important part of image processing. Region-based image segmentation method is divided into many regions according to the contents of the image。Although there is over-segmentation，studing improved algorithm can reduce the over-segmentation or choosing an effective post-processing algorithms are useful results. For example, in both Mean Shiftt and Watershed image segmentation algorithm, One can study a variety of improved algorithm to reduce over-segmentation, On the other hand, effective pretreatment can also be used to remove noise, and make the image suitable for Mean Shiftt and Watershed segmentation algorithm. MSRM is a algorithm which based on the maximum similarity of the region merging, the proposed algorithm is efficient and effective, it does not require a preset threshold and can successfully extract the single objective and mult