基于动态双阈值Canny算子对象边缘提取算法研究.doc

摘 要 边缘检测在图像理解，分析识别领域中是十分重要的研究课题，边缘检测的效果将直接影响到图像理解和识别的性能。 在图像处理领域，边缘是图像的基本特征。所谓边缘是指图像周围像素灰度有阶跃变化或屋顶状变化的像素的集合，它存在于目标和背景，目标与目标，区域与区域，基元与基元之间。边缘具有方向和幅度两个特征，沿边缘走向，像素值变化比较平缓；垂直于边缘走向，像素值变化比较剧烈，可能呈现阶跃状，也可能呈现斜坡状。 Canny算子提取算法采用二维高斯函数任一方向上的一阶方向导数为噪声滤波器，通过卷积运算对图像滤波，然后寻找滤波后图像梯度的局部极大值，以确定图像边缘。Canny算子提取算法得到的目标图像，具有噪比大和检测精度高的优点，因此得到广泛的应用。动态阈值Canny算法根据图像的具体情况而选择阈值，该方法具有更广的自适应能力，保证了图像边界提取的准确性。 关键词：边缘检测，Canny算子，动态阈值ABSTRACT Edge detection is an important topic in image understanding and identifies areas. The effect of edge detection will directly affect the image understanding and performance recognition. Edge is the most basic features of image. The so-called edge is around the pixel grayscale image with a step change or roof-like changes in the set of pixels. It exists in target and background, goals and objectives, regional and regional, unit and unit. There are two characteristic of edge, which are direction and magnitude. Along with edge, changes of pixel value are small, and in another direction changes are dramatically large. Sometimes it may shows step-like and sometimes it may be presented sloping. Canny edge detection employs 2-dimentional Gaussian filter function to eliminate noise. And then find out the maximal value of filtered image in local. There was high signal- noise ratio and accurate location of edges detected by Canny. So it is widely used in the world. Dynamic threshold of Canny method selects the threshold by every image, so it has a more adjustable and accurately. KEY WORDS: edge detection, canny operator, dynamic threshold 目 录 第1章 前 言 1 1.1 研究背景 1 1.2 Canny算子边缘提取算法的研究现状 1 1.2.1 经典算法 2 1.2.2 新兴算法 3 1.3 本文的研究内容与章节安排 5 1.4 本文的研究成果与意义 5 第2章 Canny算子边缘检测的基本理论 6 2.1 图像边缘的定义[12] 6 2.2 边缘检测的基本原理与衡量指标[13] 7 2.2.1 边缘检测的基本原理 8 2.2.2 边缘检测的衡量指标[14][15] 9 2.3 Canny算子边缘检测 9 2.3.1 Canny算子的实现步骤[16] 10 2.3.2 Canny算子的约束准则[18][19] 11 2.3.3 固定双阈值的Canny算法 12 2.4 Canny算子存在的问题及改进的方法[20] 13 2.4.1 Canny算子存在的问题 13 2.4.2 改进的Canny算法[21] 13 第3章 动态双阈值Canny算子边缘提取算法与实现 17 3.1 算法框图 17 3.