基于单应性矩阵和内容保护变形的图像拼接

Image mosaic based on homography matrix and content-preserving warping

针对图像拼接技术进行全景图像融合时出现瑕疵、畸变且不能保护图像内容的现象,提出了一种基于单应性矩阵和内容保护变形的图像拼接算法。首先选用单应性矩阵模型给定种子特征点,递增地对其相邻特征点进行分组;然后使用较合适的阈值进行分组得到尽可能多的特征点;最后通过内容保护变形算法来解决由单应性矩阵产生图像的内容发生变化的现象,使拼接后的图像不会出现物体畸变、瑕疵的现象。实验表明,该算法解决了在光照条件下出现模糊、图像内容发生变化的现象,有效地提高了图像配准的效率,能够达到无缝拼接、无视差的全景效果图。

This paper proposed an image mosaic algorithm based on homography matrix and content-preserving warping to solve the problem of image defects when image fusion was carring out.Firstly,the algorithm used the homography matrix to select the seed feature points,and grouped incrementally the adjacent feature points.Then it used the appropriate threshold to get as many feature points as possible.Finally,it used the content protection deformation algorithm to solve the phenomenon resulting in change of the image by the homography matrix,so that the spliced image would not appear object distortion,flaw,etc.Experiments show that the proposed algorithm solves the phenomenon that the fuzzy content and the image content change under the light condition,and effectively improves the efficiency of image registration,and can achieve the panorama effect with seamless splicing and no parallax.