Multi-view Matching for Unordered Image Sets, or “How Do I Organize My Holiday Snaps?”

Abstract

There has been considerable success in automated reconstruction for image sequences where small baseline algorithms can be used to establish matches across a number of images. In contrast in the case of widely separated views, methods have generally been restricted to two or three views. In this paper we investigate the problem of establishing relative viewpoints given a large number of images where no ordering information is provided. A typical application would be where images are obtained from different sources or at different times: both the viewpoint (position, orientation, scale) and lighting conditions may vary significantly over the data set. Such a problem is not fundamentally amenable to exhaustive pair wise and triplet wide baseline matching because this would be prohibitively expensive as the number of views increases. Instead, we investiate how a combination of image invariants, covariants, and multiple view relations can be used in concord to enable efficient multiple view matching. The result is a matching algorithm which is linear in the number of views. The methods are illustrated on several real image data sets. The output enables an image based technique for navigating in a 3D scene, moving from one image to whichever image is the next most appropriate.

What and Where: 3D Object Recognition with Accurate Pose

“…The computation of F between each pair of N images has N 2 complexity, thus quickly becoming prohibitively expensive with increasing N . Therefore we apply a selective approach, similar to [20], which is linear in the number of images. Image pairs are selected based on a greedy algorithm, which constructs a spanning tree on the image set.…”

3D Object Modeling and Recognition from Photographs and Image Sequences

“…The first step is to match the regions found in a pair of images. This is an instance of the wide-baseline stereo matching problem which has been well studied in the literature [3,22,24,31,35,38,44]. Any technique that generates a set of matches between affine regions in a pair of images is appropriate, including the general matching procedure (Section 2.3).…”

A Statistically Selected Part-Based Probabilistic Model for Object Recognition

“…Typically, such approaches find interest points using some operator such as [9] and then extract local image descriptors around such interest points. Several local image descriptors have been suggested and evaluated, such as Lowe's scale invariant features (SIFT) feature [11], entropy-based scale invariant features [9,6] and other local features which exhibit affine invariance such as [2,17,13]. Other approaches that model objects using local features include graph-based approaches such as [5].…”