特征点提取及视图变形技术研究

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适真。

Research on feature points extraction and view morphing

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.