为了获取宽视野的场景表示,提出了一种基于块匹配的视频图像镶嵌算法,该算法首先采用基于相位相关的块匹配方法估计出视频图像间的运动矢量场,并剔除其中由于图像噪声或运动物体的遮挡而导致外点运动矢量,然后根据图像的运动矢量场确定出图像子块之间的对应点对,进而利用得到的对应点对迭代求解图像间的变换模型参数以实现视频图像的自动镶嵌.针对真实场景的视频图像序列进行实验,获得了较好的镶嵌结果,表明了该算法的有效性.

A robust image mosaic method is proposed to construct a panoramic view of the scene from video sequence. Firstly, block matching method based on phase correlation is employed to estimate the motion vector field of video frames. After removing the outlier motion vector caused by the image noise or moving object occlusion, the corresponding point pairs between image blocks can be determined by the dominant motion vectors, and then the parameters of projective transformation model can be iteratively estimated with the obtained correspondences to implement the video mosaicing. All the work of video mosaicing is automatically completed without manual manipulation. Experimental results using video sequence with real scene have demonstrated the availability of the method.