基于图像连续表示的角点检测  

Continuous Representation of Image-Based Corner Detection

作　　者：娄联堂[1] 韦茜妤 陈佳骐 方自成 

机构地区：[1]中南民族大学数学与统计学学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2016年第2期151-156,共6页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(60975011);中南民族大学中央高校基本科研业务费专项资金资助项目(CZW15051;YZZ13003);中南民族大学研究生科研创新基金资助项目(2016sycxjj137)

摘　　要：为研究基于数字图像连续表示的角点检测方法,利用弦理论给出了离散数字图像的连续表示形式,利用快速付里叶变换实现了离散数字图像的连续表示和重建.通过计算小矩形区域的积分对直线段进行了检测.将角点的响应函数定义为平行四边形面积的大小,利用检测出的有向直线段,实现了对角点的检测.运用不同于传统的离散图像角点检测方法进行了角点检测实验,结果表明:它对图像的噪声有较强的适应性.The corner detection method based on continuous representation of digital image was studied in this paper.Firstly,the continuous representation of discrete digital image was given by string theory,Fourier transform was used to realize the continuous representation and reconstruction of discrete digital image. Then,the line segment was detected by directly calculating the integral of small rectangular area. Next,the response function of the corner was defined as the size of the parallel quadrilateral area,and the detection of corners was realized by using the detected directional line segment.Finally,the property of the algorithm was analyzed by experiments. The approach used in this paper is different from the traditional corner detection method of the discrete image. It has strong adaptability to the noise of the image.

关 键 词：数字图像连续表示 直线检测 角点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]