Registration of Challenging Image Pairs: Initialization, Estimation, and Decision

Abstract

Abstract-Our goal is an automated 2D-image-pair registration algorithm capable of aligning images taken of a wide variety of natural and man-made scenes as well as many medical images. The algorithm should handle low overlap, substantial orientation and scale differences, large illumination variations, and physical changes in the scene. An important component of this is the ability to automatically reject pairs that have no overlap or have too many differences to be aligned well. We propose a complete algorithm including techniques for initialization, for estimating transformation parameters, and for automatically deciding if an estimate is correct. Keypoints extracted and matched between images are used to generate initial similarity transform estimates, each accurate over a small region. These initial estimates are rank-ordered and tested individually in succession. Each estimate is refined using the Dual-Bootstrap ICP algorithm, driven by matching of multiscale features. A three-part decision criteria, combining measurements of alignment accuracy, stability in the estimate, and consistency in the constraints, determines whether the refined transformation estimate is accepted as correct. Experimental results on a data set of 22 challenging image pairs show that the algorithm effectively aligns 19 of the 22 pairs and rejects 99.8 percent of the misalignments that occur when all possible pairs are tried. The algorithm substantially out-performs algorithms based on keypoint matching alone.

Comparison of image registration methods for composing spectral retinal images

“…Generalized dual-bootstrap iterative closest point (GDB-ICP) [16] nds a transformation aligning two images by starting from a small area of overlap (bootstrap region) between the images and a locally stable similarity transformation. An initial transformation derived from a Scale-invariant feature transform (SIFT) descriptor match is re ned and validated by feeding edge and corner points inside a growing bootstrap region to a robust Iterative closest point (ICP) algorithm.…”

Multi-View Geometry for Camera Networks

“…We note that an excellent turnkey approach to estimating projective transformations for real images is given by the Generalized Dual-Bootstrap ICP algorithm proposed by Yang et al [15].…”

Estimating the Location of a Camera with Respect to a 3D Model

“…The approach taken here is a hypothesize-and-test strategy and an extension of the Dual-Bootstrap algorithm [13,18], originally designed for 2d-to-2d registration. A rank-ordered set of putative initial local surface-to-image mappings is generated.…”

“…These are edge-like and corner-like features, computed at multiple scales and spread throughout the images, even in low-contrast regions. Details of this computation are provided in [13,18]. These features are 1.…”

“…This requires obtaining a sufficient number of correct keypoint matches to accurately estimate and verify the 3d-to-2d projection, which can be problematic when there are substantial viewpoint, illumination and structural changes between the model and the test image. A third category of approaches, which includes ours, is based on region-growing [4,10,13,18]. In the closest work, Fraundorfer and Bischof [4] address the "kidnapped robot" problem by matching an image against a piecewise planar (indoor) scene model, using single keypoint matches to initialize and correlation to confirm.…”