吴兴蛟　吴　晟

(昆明理工大学信息工程与自动化学院　昆明　650500)



基于边缘检测的位图矢量化的实现*

吴兴蛟吴晟

(昆明理工大学信息工程与自动化学院昆明650500)

摘要论文介绍的是基于边缘检测与特征点提取的位图矢量化方法。先将位图进行RGB—HSI变换,增强图像对比度并进行边缘降噪、平滑处理,而后对图片进行阈值化处理。运用Canny算子对阈值化图像进行边缘检测,并藉此提取边缘轮廓。获得边缘轮廓后对其进行特征点求取。得到特征点集合。存储集合并在此基础上借助Matlab编写曲线处理函数,使用最小二乘法对曲线进行局部试拟合,比对SSE,找出最小值并记住当前参数,确定拟合函数。通过对所有局部曲线自动拟合技术,生成一组多项式方程,实现位图的矢将其量化。最后得出一组图像曲线。论文借助边缘检测的技术实现了位图的矢量化。

关键词边缘检测; 曲线自动拟合; 矢量化; 特征点提取; Matlab; 多项式拟合; 最小二乘法

Bitmap Vector Quantization Based on Edge Detection

WU XingjiaoWU Sheng

(School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming650500)

AbstractA bitmap vectorization methods that based on edge detection and feature point extraction is introduced in this article. Firstly, RGB-HSI transform is used to enhance the image contrast of the bitmap, to reduce the edge’s noise and smooth the edge, then applying the threshold to process the image. Using Canny operator to detect the edge of the threshold image, and to extract the edge contour. Striking the characteristic points of the edge contour, thus getting feature points set. Combining Matlab to write the curve processing function on the basis of the set storage, using the least square method to fit the curve and to compare the SSE. Then finding the minimum value and remembering the current parameters, and determining the fitting function. Through automatic fitting technique to assemble all local curves, thus generating a set of polynomial equations, then realizing the vector of bitmap and quantizing them. Finally, a set of image curves are obtained. In this paper, edge detection technology is used to achieve the vectorization of a bitmap.

Key Wordsedge detection, automatic fitting curve, vectorization, extracte of feature point, Matlab, polynomial fitting, the least square method

Class NumberTP391.41

1引言

图形(图像)通常以不同形式存储在计算机中,并用不同的方式加以描述。位图和矢量图是其中最常用的两种描述方式。位图采用像素点来描述图像。每个点用二进制数据来描述其颜色与亮度等信息,这些点是类似于点阵的离散点,多个像素的色彩组合形成位图。矢量图则是用一系列计算机指令来描述和记录图,图可以分解为一系列由点、线、面等组成的子图,矢量图所记录的是对象的几何形状、线条粗细和色彩等。图像形式可随坐标迁移[1]。……