【作者】 张洁；

【导师】 檀结庆；

【作者基本信息】 合肥工业大学 ， 计算机应用技术， 2009， 硕士

【摘要】 随着计算机技术的飞速发展,图像边缘检测已成为图像处理的重要内容,它是图像分析的基本问题,是图像分割、特征提取和图像识别的前提。本文的主要内容如下。首先,介绍了数字图像处理的概念及其应用领域、边缘检测研究的背景意义,历史现状,以及边缘检测的一些基本概念。然后,分别介绍了经典的图像边缘检测算子,如Robert算子、Sobel算子、Prewitt算子等,并通过理论分析和仿真计算比较了他们各自的优缺点及适用性。接着概述了几种新的边缘检测方法,如小波理论、数学形态学、模糊理论等。在本文的第四章里,讨论了基于线性滤波技术的边缘检测算法:Marr-Hildreth方法和Canny算法。最后,提出了一种基于各向异性扩散方程的Canny边缘检测算法。Canny边缘检测算法由于使用高斯滤波对图像进行平滑,往往使得算法的信噪比和定位精度下降,从而产生一些虚假边缘,使角点变圆。针对Canny算法所出现的问题,运用各向异性扩散方程代替高斯滤波,并对扩散后的图像做图像增强。实验结果表明,改进后的算法有效地提高了边缘检测的准确性,得到了比较理想的边缘检测效果。更多还原

【Abstract】 The image edge detection has become one of the most important parts of image processing with the development of computer technology. Image edge detection is the first step of image analysis, also the basis of image segmentation, feature extraction and image recognition. The main content of this dissertation is described as follows.Firstly, digital image processing and its applications are introduced. Then, the background, the significance and also the development status of the image edge detection technique are introduced, next to this, some basic knowledge of the image edge detection are discussed.Secondly, some classical edge detection algorithms such as Roberts, Sobel, Prewitt are discussed. Theory analysis and experiments are carried out to compare their advantages and disadvantages. Some new technology about edge detection, such as wavelet, math morphology, rough set theory, are introduced. In the fourth chapter, two kinds of edge detection algorithms based on linear filtering, i.e., Marr-Hildreth algorithm and Canny algorithm, are discussed.Finally, an improved Canny edge detection algorithm based on anisotropic diffusion equation is proposed. The canny edge detection algorithm uses Gaussian filter for smoothing the image, which may lead to lower SNR and higher edge location error, and therefore may produce false edge and corner roundness. To solve these problems, an improved edge detection algorithm is proposed by replacing Gaussian filter with anisotropic diffusion equations, and the image enhancement is carried out after diffusion. Experiment results show that the accuracy of edge detection is improved evidently, and a much better edge detection effect is obtained.更多还原