旋转扫描序列图像的投影矩阵和拼接应用

利用8参数投影矩阵(单应性矩阵)拼接旋转扫描序列图像的方法,由于矩阵估计误差的影响,配准精度不高,易导致拼接图像中的景物歪斜.针对这一问题,通过研究相机拍摄旋转扫描序列图像的透视关系,推导了一种5参数的投影矩阵,减少了单应性矩阵中8个参数之间的相关性,从而降低了投影矩阵估计误差对全景图拼接的影响,提高了配准精度.图像拼接实验结果显示,应用该5参数投影矩阵得到的全景图,能够保证场景正直,配准精度较高,视觉效果良好.

Projection Matrix of Rotary Scanning Serial Images and Its Application in Stitching

The stitching method for rotary scanning serial images using 8-parameter projection maxtrix (homography matrix), due to the influence of matrix evaluation errors, has lower matching accuracy and is easy to result in skew scenery in stitching images. To solve this problem, through studying on the perspective relationship between rotary scanning serial images taken by a digital camera, this paper deduces the 5-parameter projection matrix thus decreasing the dependency between the 8 parameters in the homography matrix, which decreases the influence of projection matrix evaluation errors on panorama stitching and increases matching accuracy. The image stitching experimental results show that, the panorama obtained using this 5-parameter projection matrix can guarantee the scene upright and has higher matching accuracy and good visual effects.