基于测地距离的图像分割讲述.docx

电子科技大学成都学院课程设计 摘 要  PAGE 27 摘 要 本文首先对数字图像处理的常见技术和发展现状以及前景作出简单的阐述，然后从原理和应用效果上对经典的图像分割方法基于测地距离的图像分割进行重点的阐述和分析，对其基本原理以及实验方法作出详细的讲解，基于测地距离的图像分割是首先对待处理图像做Gabor滤波，提取滤波后各个子带图像通道的窗口能量特征；然后，对目标和背景的标记像素点各通道特征单独建立高斯概率模型，计算得到测试样本在各个通道上分属于两个高斯的概率，并加权各通道得到各像素点属于前景和背景的权值概率矩阵；最后，把权值概率矩阵看作有向带权图，图像列化坐标值作为顶点值，权值概率矩阵中对应的权值作为边的权，将图像分割问题可以转化为所有像素点到目标和背景的标记像素点的最短路径问题，根据本章所提出的计算图像像素点到目标与背景像素点的最小测地距离方法来确定各个像素点的类别，完成图像分割任务。最后，在图像分割的实施过程中必须运用的MATLAB软件进行简单的理解，熟悉运用环境和使用方法。 关键字：测地距离，图像分割，目标图像，背景图像 ABSTRACT 目 录  PAGE II ABSTRACT Based on digital image processing technology and development situation and Prospect of common makes a simple exposition, and then from the theory and the application effect to the classic image segmentation methods This article from the principle and the application effect to the classic image segmentation method based on geodesic distance image segmentation to carry on the elaboration and the analysis.Image segmentation based on geodesic distance is processed image Gabor filtering, extracting filter after each sub-band image channel window energy characteristics;Then, the target and background labeling pixels each channel features separate establishment of Gauss probability model, calculated the test samples in each channel belong to two Gauss probability, and weighted each channel of each pixel belongs to the background and foreground probability weights matrix;Finally, the probability weights matrix as a weighted directed graph, image column of coordinate value as a vertex value, weighted probability matrix of the corresponding weight as a side of the right, the image segmentation problem can be transformed into all the pixels into target and background marker pixel points of the shortest path problem, according to this chapter proposed calculation of image pixels to the target and background pixels minimal geodesic distance method to determine each pixel point category, to complete the task of image segmentation.Finally, in the image segmentation of the impleme