基于多尺度分形理论的图像边缘提取算法及其在医学图像中的应用毕业设计论文.doc

摘 要 自从Mandelbrot提出分形的概念以来，分形学已发展成为一门横跨自然科学和社会科学各研究领域的新兴学科。分形学的基本思想是：客观事物具有自相似的层次结构，局部与整体在形态、功能、信息、时间、空间等方面具有统计意义上的相似性，即自相似性。自相似性原理的引入使分形理论成为研究和处理自然与工程中不规则图形的有力工具。广义而言，任何物体表面都可以看作是由某种纹理特征的表面构成，故任何图像都包含了若干种纹理区域的灰度表面。在这些不同纹理灰度表面之间灰度起伏变化显著，外在就表现为边缘。大多数纹理图像都可以用分形模型进行描述，而纹理特征的变化包含了图像的边缘信息。目前运用分形理论进行图像边缘检测已成为分形图形学的一种重要应用。 本文在图像分形模型—分数布朗随机场的基础上，通过分析图像的分形参数，提出一种新的边缘检测特征度量，并根据边缘分形特征计算出自适应阈值，实现图像的边缘检测，并且将利用该算法提取的边缘图像与用传统的微分算子提取的边缘图像进行了比较以证明本文提出的方法是一种更有效的边缘检测算法。在此基础上，本人提出了两种改进算法，改进算法可以大大节省程序运行的时间，提高了程序的执行效率，同时改进算法也可以达到很好的边缘检测效果。对此算法在实际中的应用本文用专门的篇章进行了详细的讨论。 本文首先介绍了分形产生的背景以及近年来发展的状况，在此基础上提出了一种基于计算纹理图像的分形特征进行边缘检测的算法，证明了该算法优于传统的微分算子的方法，进而又提出了两种改进算法，即分形与图像扫描技术相结合的算法。最后将该分形算法应用于医学图像处理中，取得了良好的效果，并提出了一些改进建议。 关键词：分形理论、分形维数、分数布朗运动、毯子算法 Abstract Since Mandelbrot created the concept of fractal, the Fractal had become a new subject crossing the natural science and social science. The basic idea of the Fractal is that impersonality substance has the self-similar arrange structure, the part and the whole are statistically similar in the form, function, information, time and space, it is also called self-similarity. The introduction of the principle of self-similar let the fractal theory become a power tool of studying and manipulating the anomaly image in nature and project. In a broad sense, the surface of any object can be seen as being made of surfaces of some texture character, any image includes gray level surfaces which is made of some kinds of texture. There are significant changes of gray level between different kinds of texture surfaces, and it is called edge. Most texture image can be described by a fractal model, the changes of texture character include the information of edge of the image. Now using the fractal method to detect the edge of image has been a important application of fractal figure subject. In this paper, on the base of image fractal model-Fraction Brownian Motion, the author will analysis the fractal parameter of the image to report a new edge detection character