基于单应性矩阵的图像拼接方法的对比分析

Comparisons and Analysis of Homography-based Image Stitching Methods

图像拼接是将一系列图像结合成一个宽场景的图像。本文简述了图像拼接的一般流程,重点介绍了三种基于单应性(Homograohy)矩阵的图像拼接方法。文中利用三种图像拼接方法对两组图像进行拼接实验。实验结果表明:如果两幅图像在同一个平面场景或者拍照中心相同,用一个基本的单应性矩阵能完成拼接;如果两张幅像之间存在除了旋转之外的变换或者不在同一平面场景,由于单应性矩阵无法处理视差,仅使用单应性矩阵来拼接得到的结果将出现鬼影或错位。

Image stitching is a combination of a series of images into a wide view image. The paper briefly describes the general process of image stitching, mainly introduces three image stitching method based on homography matrix. In this paper, the three methods are used to stitching two sets of images. The result shows that： if the two images scenes are effectively planar or have same camera center, using a basic homography matrix can complete stitching; if there is another transformation except rotation between two images or they don ＇t in a same plane scene, as the homography matrix cannot account for parallax, stitching by homoagrapy matrix alone will result in ghosting and dislocation.