Correspondence Expansion for Wide Baseline Stereo

Abstract

We present a new method for generating large numbers of accurate point correspondences between two wide baseline images. This is important for structure-from-motion algorithms, which rely on many correct matches to reduce error in the derived geometric structure. Given a small initial correspondence set we iteratively expand the set with nearby points exhibiting strong affine correlation, and then we constrain the set to an epipolar geometry using RANSAC. A key point to our algorithm is to allow a high error tolerance in the constraint, allowing the correspondence set to expand into many areas of an image before applying a lower error tolerance constraint. We show that this method successfully expands a small set of initial matches, and we demonstrate it on a variety of image pairs.

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

“…Region models more sophisticated than our rectangular model could be developed-e.g., ones encompassing only the points where the trace of the transfer error covariance matrix is below a threshold-but the rectangular shape has proven sufficient for all our experiments and has not been the cause of an algorithm failure. Note that other region growth techniques have recently been proposed in the literature [17], [47]. Ours differs in that its growth is monotonic and is controlled by the uncertainty in the transformation estimate.…”

Multimodal image registration based on edges and junctions

“…Bay et al 9 extracted edges using a Canny detector 10 and matched straight line segments between two uncalibrated wide-baseline images without assuming epipolar geometry to be known beforehand. Given a small set of initial correspondences, Steele et al 11 generated a large number of point correspondences between two wide baseline images. These points can function as extrinsic control points, 1 and thus the spatial mapping can be determined directly through aligning those control points, not involving any optimization methods.…”

“…Consequently, it is not an easy task to design or choose a highly-efficient optimizer, although it has a large influence on the registration results. (2) If a sufficient number of correct point correspondences are generated by the presented algorithms, 11,14 proper local transformations can be established to account for local misalignments and hence provide more accurate registration results. Additionally, it does not require segmenting images into small 'even' regions, with each aligned by a set of transformation parameters.…”

Wide Baseline Matching Using Support Vector Regression

“…New matches were searched guided by local affine transformations, and a sidedness topological constraint for triple matches was used to remove bad matches.Tuytelaars and Gool gave two local constraints [5], and the matches with sufficient other matches obeying the same local constraints were reserved. Steele and Egbert [9] expanded the matches guided by local similarity transformation estimated previously, and constrained the matches by epipolar geometry with RANSAC estimation using a high inlier error tolerance. Using global geometric relationships-the basic matrix [2] or local geometric relationslocal affine transformation to guide the match [2,9].…”

“…Steele and Egbert [9] expanded the matches guided by local similarity transformation estimated previously, and constrained the matches by epipolar geometry with RANSAC estimation using a high inlier error tolerance. Using global geometric relationships-the basic matrix [2] or local geometric relationslocal affine transformation to guide the match [2,9]. However,the fundamental matrix just provides a bilinear constraint, if there is no adequate match, then it can not be a better estimate.Local geometric relations can provide prediction image, but it is very sensitive.…”