A Robust Image Mosaic Algorithm

Image mosaic is important in the filed of computer vision. In this paper, SUSAN corner detector is used to extract feature point, and the phase correlation technique is used to roughly estimate translational parameters of two images. The wrongly matched points are deleted with RANSAC algorithm after the initial matching. At the same time, fundamental matrix and homography matrix are estimated robustly based on the epipolar and homography constraints. The linear weighted transition method is employed in the process of image fusion. Experiments show the robustness and precision of this algorithm.