为了快速且直观地对相机弱标定 (即相机内参数已知而外参数未知) 情况下的立体图像对进行欧氏极线校正, 提出了一套完整的欧氏极线校正框架。首先, 针对两视图对应相机的位姿信息估计问题, 提出一种新的不依赖矩阵奇异值分解运算的本质矩阵分解和唯一解确定算法。然后, 对传统欧氏极线校正问题进行推广, 并利用估计出的相机位姿构造出具有明确几何意义的无穷单应变换对, 进而将其施加到立体图像对来完成欧氏极线校正。最后, 利用8组SYNTIM立体图像对验证了所提出校正框架的正确性和校正精度。实验结果表明: 该欧氏极线校正框架不仅形式直观, 计算简单, 易于实现, 而且具有明确的几何意义。与其他欧氏极线校正方法比较, 提出的方法能够获得更高的校正精度, 且更好地消除了立体图像对的水平或垂直视差。

To quickly and intuitively finish the Euclidean epipolar rectification of weakly calibrated stereo pairs(namely,the camera intrinsic parameters are known and the camera external parameters are unknown), an integral Euclidean epipolar rectification frame was proposed. Firstly, a novel essential matrix decomposition and unique solution determination algorithm independent on the Singular Value Decomposition (SVD) operation was proposed to determine the position and orientation information between two cameras. Then, the problem of traditional Euclidean epipolar rectification was generalized, and the infinite homography matrixes were constructed by the estimated camera position and orientation to finish the rectification process of stereo pairs. Finally, the eight groups of SYNTIM images were used to validate the correctness and accuracy of the rectification frame. Experimental results indicate that the proposed frame has the advantages of intuitive form, low computational complexity and easy implementation, as well as the explicit geometrical meanings. As comparison with traditional Euclidean epipolar rectification method, it can well remove the horizontal or vertical disparities of stereo pairs, and can offer higher rectification accuracy.