Homography-Based Planar Object Reconstruction From Uncalibrated Images

摘要

We address the 3D projective structure determination of planar objects using the planar homographies derived from the plane features in the images. The epipolar geometry of the image pair can be represented by a planar homography, an epipole, and a specified reference plane coefficient vector. Next, we specify the 3D plane coefficient vector of any plane visible in each image pair with respect to a reference plane if its planar homography is found. Finally, the reconstruction results obtained in individual projective spaces are integrated under a common projective space. We avoid using the corresponding points to estimate the required coordinate transformation between two involved projective spaces; instead, we use the plane-based information to derive the coordinate transformation matrix. To evaluate the performance of our method, we apply our method to the synthetic and real images. All the results indicate the method works successfully.