Epipolar Geometry Estimation: General Settings & Special Conditions

In the talk I will review some of my works on Epipolar Geometry Estimation. I will start with the BEEM algorithm which puts RANSAC within the general framework of search/optimization methods. Suggestions for the various components of the algorithm will be given.


In the second part of the talk I will present our recent work on matching images of urban scenes with repetitive structures. General purpose algorithms usually fail such image pairs. We deal with two common cases.In the first case it is assumed that the building façade has a grid structure. In this case we can automatically recover most of the matching parameters. We also deal with the more general case in which repeated objects are partially organized horizontally or vertically. In both cases we are able to match the building façade and then recover the epipole. The algorithm has been run successfully on a large number of image pairs from the ZuBuD database for which state of the art general purpose algorithms fail.


Finally we will describe our new algorithm on Epipolar Geometry Estimation Using Noisy Pose Priors.Smartphones are equipped with sensors which are able to estimate the internal and external calibration parameters of the images they take. A similar situation occurs when images are taken by a robot equipped with an IMU.It is therefore natural to study how these parameters can be used for Epipolar geometry estimation. The main challenge is that these parameters are quite noisy and therefore naïve methods cannot be used.


We introduce SOREPP, a novel estimation algorithm designed to exploit pose priors naturally. It sparsely samples the pose space around the measured pose and for a few promising candidates applies a robust optimization procedure. It uses all the putative correspondences simultaneously, even though many of them are outliers, yielding a very efficient algorithm whose runtime is independent of the inlier fractions.SOREPP was extensively tested on synthetic data and on hundreds of real image pairs taken by a smartphone. Its ability to handle challenging scenarios with extremely low inlier fractions of less than 10% was demonstrated as was its ability to handle close cameras.