Fast Image-matching Technique Robust to Rotation in Spherical Images
Chanho Moon, Seong-Won Lee · IEIE Transactions on Smart Processing and Computing · 2020
There are many studies on 3D reconstruction using image-matching algorithms for 360-degree spherical images. However, it is difficult to apply conventional planar-based matching algorithms because a spherical image is represented in 3D space with a geodesic grid. In this study, a fast image-matching algorithm especially robust to rotation in spherical images is proposed. With interleaving and hashing techniques, memory waste is reduced and memory access efficiency is improved in mapping a 2D flatten icosahedron corresponding to a 3D icosphere. In addition, a modified Features from Accelerated Segment Test–based matching technique is proposed to improve robust matching performance. The proposed matching technique uses a descriptor with directional information order to endure rotation of images.