基于MSER与ASIFT的鱼眼镜头图像立体匹配方法

Stereo Matching Method of Fisheye Lens Image Based on MSER and ASIFT

介绍图像立体匹配原理、最稳定极值区域MSER和特征描述算子ASIFT。MSER+SIFT、MSER+SURF和MSER+ASIFT算法均可以应用在未经畸变矫正的鱼眼图像立体匹配中。实验结果表明,MSER+ASIFT算法对大视角变换和畸变较大的图像匹配具有很好的鲁棒性,虽然运行速度上没有明显优势,但相比其他算法而言,该算法匹配结果最准确。

The working principle of image stereo matching, the MSER （maximally stable extremal region） and the ASIFT feature description operator were introduced. The algorithms of MSER ＋ SIFT, MSER ＋ SURF and MSER ＋ ASIFT can be applied to the stereo matching of fisheye images without distortion correction. The ex- perimental results show that, as for the image matching with great visual changes and big distortion, the MSER ＋ ASIFT algorithm has better robustness although there is no obvious advantage in its running speed; com- pared with other algorithms, this algorithm matches the results most accurately.