We present a novel algorithm (which we call “Frag-Track”) for tracking an object in a video sequence. The template object is represented by multiple image fragments or patches. The patches are arbitrary and are not based on an object model (in contrast with traditional use of modelbased parts e.g. limbs and torso in human tracking). Every patch votes on the possible positions and scales of the object in the current frame, by comparing its histogram with the corresponding image patch histogram. We then minimize a robust statistic in order to combine the vote maps of the multiple patches. A key tool enabling the application of our algorithm to tracking is the integral histogram data structure [18]. Its use allows to extract histograms of multiple rectangular regions in the image in a very efficient manner. Our algorithm overcomes several difficulties which cannot be handled by traditional histogram-based algorithms [8, 6]. First, by robustly combining multiple patch votes, we are able to handle partial occlusions or pose change. Second, the geometric relations between the template patches allow us to take into account the spatial distribution of the pixel intensities- information which is lost in traditional histogram-based algorithms. Third, as noted by [18], tracking large targets has the same computational cost as tracking small targets. We present extensive experimental results on challenging sequences, which demonstrate the robust tracking achieved by our algorithm (even with the use of only gray-scale (noncolor) information). 1.

We propose a new method to estimate multiple rigid motions from noisy 3D point correspondences in the presence of outliers. The method does not require prior specification of number of motion groups and estimates all the motion parameters simultaneously. We start with generating samples from the rigid motion distribution. The motion parameters are then estimated via mode finding operations on the sampled distribution. Since rigid motions do not lie on a vector space, classical statistical methods can not be used for mode finding. We develop a mean shift algorithm which estimates modes of the sampled distribution using the Lie group structure of the rigid motions. We also show that proposed mean shift algorithm is general and can be applied to any distribution having a matrix Lie group structure. Experimental results on synthetic and real image data demonstrate the superior performance of the algorithm. 1.

Random Sample Consensus (RANSAC) is the most widely used robust regression algorithm in computer vision. However, RANSAC has a few drawbacks which make it difficult to use for practical applications. Some of these problems have been addressed through improved sampling algorithms or better cost functions, but an important difficulty still remains. The algorithm is not user independent, and requires knowledge of the scale of the inlier noise. We propose a new robust regression algorithm, the projection based M-estimator (pbM). The pbM algorithm is derived by building a connection to the theory of kernel density estimation and this leads to an improved cost function, which gives better performance. Furthermore, pbM is user independent and does not require any knowledge of the scale of noise corrupting the inliers. We propose a general framework for the pbM algorithm which can handle heteroscedastic data and multiple linear constraints on each data point through the use of Grassmann manifold theory. The performance of pbM is compared with RANSAC and M-Estimator Sample Consensus (MSAC) on various real problems. It is shown that pbM gives better results than RANSAC and MSAC in spite of being user independent.

ABSTRACT: Change detection is important for an up-to-date GIS database. The ever improving spatial, spectral and temporal resolution of satellite imagery allows for reliable detection and characterization of even more details of the changed patterns with higher accuracy. The quality of registration of the involved imagery is the key factor that dictates the validity and the reliability of the change detection results. The fact that the change detection process usually involves multi-spectral and/or multi-resolution imagery captured at different times and from different sensors emphasises the issue of development of a robust registration procedure that can handle these types of images. This paper introduces a new approach for automated image registration based on a hierarchical image matching strategy. After feature point extraction, the method uses the similarity of the grey levels to find the candidates of the homologous points across the images. To increase success rate and reliability, and reduce computational complexity, a hierarchical image pyramid has been used. Matching then starts from the highest pyramid level with the results being the approximation of the subsequent lower level. The algorithm also uses contextual information to achieve locally consistent matches. The method has been implemented and tested using various remote sensing imagery including IKONOS and QuickBird data over test sites in Melbourne, Australia and Thimphu, Bhutan. The results are promising and reveal the potential for operational automated image registration in the process of change detection. 1.

Abstract. We propose a solution to the problem of robust subspace estimation using the projection based M-estimator. The new method handles more outliers than inliers, does not require a user defined scale of the noise affecting the inliers, handles noncentered data and nonorthogonal subspaces. Other robust methods like RANSAC, use an input for the scale, while methods for subspace segmentation, like GPCA, are not robust. Synthetic data and three real cases of multibody factorization show the superiority of our method, in spite of user independence. 1

In this paper we propose a strategy for the orientation and auto-calibration of wide-baseline image sequences. Our particular contribution lies in demonstrating, that by means of robust least-squares adjustment in the form of bundle adjustment as well as least-squares matching (LSM), one can obtain highly precise and reliable results. To deal with large image sizes, we make use of image pyramids. We do not need approximate values, neither for orientation nor calibration, because we use direct solutions and robust algorithms, particularly fundamental matrices F, trifocal tensors � , random sample consensus (RANSAC), and auto-calibration based on the image of the dual absolute quadric. We describe our strategy from end to end, and demonstrate its potential by means of examples, showing also one way for evaluation. The latter is based on imaging a cylindrical object (advertisement column), taking the last to be the first image, but without employing the closedness constraint. We finally summarize our findings and point to further directions of research. 1

We propose a new robust estimator for camera pose estimation based on a recently developed nonlinear mean shift algorithm. This allows us to treat pose estimation as a clustering problem in the presence of outliers. We compare our method to RANSAC, which is the standard robust estimator for computer vision problems. We also show that under fairly general assumptions our method is provably better than RANSAC. Synthetic and real examples to support our claims are provided. 1.

Abstract. Existing multi-kernel tracking methods are based on a forwards additive motion model formulation. However this approach suffers from the need to estimate an update matrix for each iteration. This paper presents a general framework that extends the existing approach and that allows to introduce a new inverse compositional formulation which shifts the computation of the update matrix to a one time initialisation step. The proposed approach thus reduces the computational complexity of each iteration, compared to the existing forwards approach. The approaches are compared both in terms of algorithmic complexity and quality of the estimation. 1

Abstract. Structure-from-motion (SfM) is an important computer vision problem and largely relies on the quality of feature tracking. In image sequences, if disjointed tracks caused by objects moving in and out of the view, occasional occlusion, or image noise, are not handled well, the corresponding SfM could be significantly affected. In this paper, we address the non-consecutive feature point tracking problem and propose an effective method to match interrupted tracks. Our framework consists of steps of solving the feature ‘dropout ’ problem when indistinctive structures, noise or even large image distortion exist, and of rapidly recognizing and joining common features located in different subsequences. Experimental results on several challenging and large-scale video sets show that our method notably improves SfM. 1

Abstract: GIS databases need to undergo frequent updating due to rapid changes in the physical environment. Satellite imagery, with its recently enhanced spatial, spectral and temporal resolution, allows for accurate and reliable detection and characterization of patterns of change. A key factor that dictates the validity and reliability of change detection results is the quality of image registration. There is a need to develop robust registration procedures for multi-temporal, multi-spectral and/or multi-resolution imagery captured from different sensors to facilitate the change detection process. This paper discusses an approach for automated image registration based on a hierarchical image matching strategy. Following an initial feature point extraction phase, the method uses the similarity of grey levels to find homologous point candidates across the images. A hierarchical image pyramid is used to enhance both the success rate and reliability of matching while reducing the computational complexity. The adopted approach also uses contextual information to achieve locally consistent matches. Implementation and testing of the proposed method has involved imagery from different