Catálogo de publicaciones - libros

The object of this paper is to present a formulation for the segmentation and restoration problem using flexible models with a robust regularized network (RRN). A two-steps iterative algorithm is presented. In the first step an approximation of the classification is computed by using a local minimization algorithm, and in the second step the parameters of the RRN are updated. The use of robust potentials is motivated by (a) classification errors that can result from the use of local minimizer algorithms in the implementation, and (b) the need to adapt the RN using local image gradient information to improve fidelity of the model to the data.

Principal Component Analysis (PCA) is one of the most popular techniques for dimensionality reduction of multivariate data points with application areas covering many branches of science. However, conventional PCA handles the multivariate data in a discrete manner only, i.e., the covariance matrix represents only sample data points rather than higher-order data representations.

In this paper we extend conventional PCA by proposing techniques for constructing the covariance matrix of uniformly sampled continuous regions in parameter space. These regions include polytops defined by convex combinations of sample data, and polyhedral regions defined by intersection of half spaces. The applications of these ideas in practice are simple and shown to be very effective in providing much superior generalization properties than conventional PCA for appearance-based recognition applications.

This paper presents a geometric approach to recognizing smooth objects from their outlines. We define a signature function that associates feature vectors with objects and baselines connecting pairs of possible viewpoints. Feature vectors, which can be projective, affine, or Euclidean, are computed using the planes that pass through a fixed baseline and are also tangent to the object’s surface. In the proposed framework, matching a test outline to a set of training outlines is equivalent to finding intersections in feature space between the images of the training and the test signature functions. The paper presents experimental results for the case of internally calibrated perspective cameras, where the feature vectors are angles between epipolar tangent planes.

The problem we consider in this paper is to take a single two-dimensional image containing a human body, locate the joint positions, and use these to estimate the body configuration and pose in three-dimensional space. The basic approach is to store a number of exemplar 2D views of the human body in a variety of different configurations and viewpoints with respect to the camera. On each of these stored views, the locations of the body joints (left elbow, right knee, etc.) are manually marked and labelled for future use. The test shape is then matched to each stored view, using the technique of shape context matching in conjunction with a kinematic chain-based deformation model. Assuming that there is a stored view sufficiently similar in configuration and pose, the correspondence process will succeed. The locations of the body joints are then transferred from the exemplar view to the test shape. Given the joint locations, the 3D body configuration and pose are then estimated. We can apply this technique to video by treating each frame independently - tracking just becomes repeated recognition! We present results on a variety of datasets.

This paper presents a method for incorporating temporal information in a video sequence for the task of human recognition. A time series state space model, parameterized by a and a , is proposed to simultaneously characterize the kinematics and identity. Two (SIS) methods, a brute-force version and an efficient version, are developed to provide numerical solutions to the model. The joint distribution of both state vector and identity variable is estimated at each time instant and then propagated to the next time instant. over the state vector yields a robust estimate of the posterior distribution of the identity variable. Due to the propagation of identity and kinematics, a in posterior probability of the identity variable is achieved to give improved recognition. This evolving behavior is characterized using changes in . The effectiveness of this approach is illustrated using experimental results on low resolution face data and upper body data.

The problem of pose estimation arises in many areas of computer vision, including object recognition, object tracking, site inspection and updating, and autonomous navigation using scene models. We present a new algorithm, called , for determining the pose of a 3D object from a single 2D image in the case that correspondences between model points and image points are unknown. The algorithm combines Gold’s iterative SoftAssign algorithm [, for computing correspondences and DeMenthon’s iterative POSIT algorithm [] for computing object pose under a full-perspective camera model. Our algorithm, unlike most previous algorithms for this problem, have to hypothesize small sets of matches and then verify the remaining image points. Instead, possible matches are treated identically throughout the search for an optimal pose. The performance of the algorithm is extensively evaluated in Monte Carlo simulations on synthetic data under a variety of levels of clutter, occlusion, and image noise. These tests show that the algorithm performs well in a variety of difficult scenarios, and empirical evidence suggests that the algorithm has a run-time complexity that is better than previous methods by a factor equal to the number of image points. The algorithm is being applied to the practical problem of autonomous vehicle navigation in a city through registration of a 3D architectural models of buildings to images obtained from an on-board camera.

We derive a pseudo-metric for weighted point sets. There are numerous situations, for example in the shape description domain, where the individual points in a feature point set have an associated attribute, a weight. A distance function that incorporates this extra information apart from the points’ position can be very useful for matching and retrieval purposes. There are two main approaches to do this. One approach is to interpret the point sets as fuzzy sets. However, a distance measure for fuzzy sets that is a metric, invariant under rigid motion and respects scaling of the underlying ground distance, does not exist. In addition, a Hausdorff-like pseudo-metric fails to differentiate between fuzzy sets with arbitrarily different maximum membership values. The other approach is the Earth Mover’s Distance. However, for sets of unequal total weights, it gives zero distance for arbitrarily different sets, and does not obey the triangle inequality. In this paper we derive a distance measure, based on weight transportation, that is invariant under rigid motion, respects scaling, and obeys the triangle inequality, so that it can be used in efficient database searching. Moreover, our pseudo-metric identifies only weight-scaled versions of the same set. We demonstrate its potential use by testing it on two different collections, one of company logos and another one of fish contours.

This paper examines issues arising in applying a previously developed edit-distance shock graph matching technique to indexing into shape databases. This approach compares the shock graph topology and attributes to produce a similarity , and results in 100% recognition rate in querying a database of approximately 200 shapes. However, indexing into a significantly larger database is faced with both the lack of a suitable database, and more significantly with the expense related to computing the metric. We have thus () gathered shapes from a variety of sources to create a database of over 1000 shapes from forty categories as a stage towards developing an approach for indexing into a much larger database; () developed a approximate similarly measure which relies on the shock graph topology and a very coarse sampling of link attributes. We show that this is a good first-order approximation of the similarly metric and is two orders of magnitude more efficient to compute. An interesting outcome of using this efficient but approximate similarity measure is that the approximation naturally demands a notion of to give high precision; () developed an exemplar-based indexing scheme which discards a large number of non-matching shapes solely based on distance to exemplars, coarse scale representatives of each category. The use of a coarse-scale matching measure in conjunction with a coarse-scale sampling of the database leads to a significant reduction in the computational effort without discarding correct matches, thus paving the way for indexing into databases of tens of thousands of shapes.

Eigensegments combine image segmentation and Principal Component Analysis (PCA) to obtain a spatio-temporal decomposition of an ensemble of images. The image plane is decomposed into temporally correlated regions. Each region is independently decomposed using PCA. Thus, each image is modeled by several low-dimensional segment-spaces, instead of a single high-dimensional image-space. Experiments show the proposed method gives better classification results, gives smaller reconstruction errors, can handle local changes in appearance and is faster to compute. Results for faces and vehicles are shown.

We present a framework for representing and matching multi-scale, qualitative feature hierarchies. The coarse shape of an object is captured by a set of blobs and ridges, representing compact and elongated parts of an object. These parts, in turn, map to nodes in a directed acyclic graph, in which parent/child edges represent feature overlap, sibling edges join nodes with shared parents, and all edges encode geometric relations between the features. Given two feature hierarchies, represented as directed acyclic graphs, we present an algorithm for computing both similarity and node correspondence in the presence of noise and occlusion. Similarity, in turn, is a function of structural similarity, contextual similarity (geometric relations among neighboring nodes), and node contents similarity. Moreover, the weights of these components can be varied on a , allowing a graph-based model to effectively parameterize the saliency of its constraints. We demonstrate the approach on two domains: gesture recognition and face detection.