摘要
针对现有基于特征的拼接算法在特征单一的场景中存在配准精度和效率方面的不足,提出了一种新的拼接算法,巧妙地将摄像机标定与相位相关的方法相结合,通过摄像机标定求出摄像机内外参数,对摄像机输入的视频图像进行矫正,从而使得两幅待匹配的图像间只具有平移参量。再通过相位相关算法求取高精度的平移参数,最后进行图像融合处理。拼接过程中,不需要提取拼接图像的特征,减少了图像配准过程中算法对于场景特征的依赖性。实验结果证明,该方法在场景特征单一的条件下能够实现稳定高效的视频拼接。
In single scene the existing feature-based mosaicking algorithm registration accuracy and efficiency,a novel stitching algorithm is presented combining camera calibration and phase correlation. Firstly, computing interior and exterior parameters of the camera by calibration, the input video images could be corrected by the camera parameters, so that the two images only have translation parameters. Then using phase correlation algorithm to compute the high accuracy translation parameters,the final step is fusion processing. During the splicing processing doesn＇ t require extraction of image feature,and greatly reduces the dependence of scene features. The experimental results show that the method can achieve a stable and efficient video stitching even in the single scene.
出处
《电视技术》
北大核心
2013年第17期151-154,158,共5页
Video Engineering
基金
国家科技支撑计划项目(2012BAH20B01)
广西自然科学基金项目(2010GXNSFC013014)
关键词
视频拼接
摄像机标定
傅里叶变换
相位相关
video mosaic
camera calibration
Fourier transform
phase correlation
