特征点提取及视图变形技术研究  

Research on feature points extraction and view morphing

作　　者：华顺刚[1] 逄岭[1] 欧宗瑛[1] 

机构地区：[1]大连理工大学机械工程学院CAD&CG研究所,辽宁大连116024

出　　处：《大连理工大学学报》2004年第5期674-680,共7页Journal of Dalian University of Technology

摘　　要：介绍了图像变形和视图变形技术,在此基础上提出了一种改进的视图变形方法.利用样条小波变换进行边缘检测,根据模极大值原理,获得单像素的轮廓边缘;针对传统的霍夫(Hough)变换的局限性,对其进行改进,并将改进后的Hough变换方法,用于从单像素边缘图像中提取直线的特征点,即端点中.在原始图像中,载入这些特征点,交互地选取其中所需的特征点,连接形成特征线段,然后使用图像变形方法,生成从起始图像到终止图像的期间过渡图像,实现两幅图像的渐变.该方法提高了特征点选取的准确性和效率,使中间过渡图像更加自然逼真.On the basis of the image morphing and view morphing, an improved view morphing method is presented. The single pixel edges can be obtained from original image by using quadric spline wavelet transform and model maximum principle, feature points (beginning and ending) of line segments can be found from the edge images by using improved Hough transform. With the developed interface, both original images and feature points from which the desired points exactly selected are loaded, and corresponding lines between two images are defined. As a result, each intermediate frame can be created by using view morphing technique. Through the improved method, the accurate corresponding points and a convincing sequence of transition images between a pair of original images can be obtained. The accuracy and efficiency of feature points extraction are promoted.

关 键 词：视图变形 特征点提取 图像变形 边缘图像 边缘检测 HOUGH变换 交互 线段 直线 方法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] G633[自动化与计算机技术—计算机科学与技术]