Many different low-level feature detectors exist and it is widely agreed that the evaluation of detectors is important. In this paper we introduce two evaluation criteria for interest points: repeatability rate and information content. Repeatability rate evaluates the geometric stability under different transformations. Information content measures the distinctiveness of features. Different interest point detectors are compared using these two criteria. We determine which detector gives the best results and show that it satisfies the criteria well.

Abstract — We study the problem of detecting objects in still, grayscale images. Our primary focus is development of a learning-based approach to the problem, that makes use of a sparse, part-based representation. A vocabulary of distinctive object parts is automatically constructed from a set of sample images of the object class of interest; images are then represented using parts from this vocabulary, together with spatial relations observed among the parts. Based on this representation, a learning algorithm is used to automatically learn to detect instances of the object class in new images. The approach can be applied to any object with distinguishable parts in a relatively fixed spatial configuration; it is evaluated here on difficult sets of real-world images containing side views of cars, and is seen to successfully detect objects in varying conditions amidst background clutter and mild occlusion. In evaluating object detection approaches, several important methodological issues arise that have not been satisfactorily addressed in previous work. A secondary focus of this paper is to highlight these issues and to develop rigorous evaluation standards for the object detection problem. A critical evaluation of our approach under the proposed standards is presented.

This paper describes a new approach to low level image processing; in particular, edge and corner detection and structure preserving noise reduction.

Abstract-An algorithm for matching images of real world scenes is presented. The matching is a specification of the geometrical disparity between the images and may be used to partially reconstruct the threedimensional structure of the scene. Sets of candidate matching points are selected independently in each image. These points are the locations of small, distinct features which are likely to be detectable in both images. An initial network of possible matches between the two sets of candidates is constructed. Each possible match specifies a possible disparity of a candidate point in a selected reference image. An initial estimate of the probability of each possible disparity is made, based on the similarity of subimages surrounding the points. These estimates are iteratively improved by a relaxation labeling technique making use of the local continuity property of disparity that is a consequence of the continuity of real world surfaces. The algorithm is effective for binocular parallax, motion parallax, and object motion. It quickly converges to good estimates of disparity, which reflect the spatial organization of the scene. Index Terms-Disparity, matching, motion, relaxation labeling, scene analysis, stereo.

Corners and vertices are strong and useful features in Computer Vision for scene analysis, stereo matching and motion analysis. This paper deals with the development of a computational approach to these important features. We consider first a corner model and study analytically its behavior once it has been smoothed using the well-known Gaussian filter. This allows us to clarify the behavior of some well known cornerness measure based approaches used to detect these points of interest. Most of these classical approaches appear to detect points that do not correspond to the exact position of the corner. A new scale-space based approach that combines useful properties from the Laplacian and Beaudet's measure [Bea78] is then proposed in order to correct and detect exactly the corner position. An extension of this approach is then developed to solve the problem of trihedral vertex characterization and detection. In particular, it is shown that a trihedral vertex has two elliptic maxima on ...

Motion perception and interpretation plays an important role in the human visual system. It helps us recognize different objects and their motion in a scene, infer their relative depth, their rigidity, etc. In psychology, this process has been studied extensively by Johansson using moving light displays (MLDs). MLDs consist of bright spots attached to the joints of an actor dressed in black, and moving in front of a dark background. The collection of spots carry only 2D information and no structural information, since they are not connected. A set of static spots remained meaningless to observers, while their relative movement created a vivid impression of a person walking, running, dancing, etc. The gender of a person, and even the gait of a friend can be recognized based solely on the motion of those spots. There are two theories about the interpretation of MLD type stimuli, from a psychology point of view. In the first, people use motion information in the MLD to recover t...

Computational models of the human stereo system' can provide insight into general information processing constraints that apply to any stereo system, either artificial or biological. In 1977, Marr and Poggio proposed one such computational model, that was characterized as matching certain feature points in difference-of-Gaussian filtered images, and using the information obtained by matching coarser resolution representations to restrict the search'space for matching finer resolution representations. An implementation of the algorithm and'its testing on a range of images was reported in 1980. Since then a number of psychophysical experiments have suggested possible refinements to the model and modifications to the algorithm. As well, recent computational experiments applying the algorithm to a variety of natural images, especially aerial photographs, have led to a number of modifications. In this article, we present a version of the Marr-Poggio-Gfimson algorithm that embodies these modifications and illustrate its performance on a series of natural images.

A stereo vision system that is required to support high-level object based tasks in a tele-operated environment is described. Stereo vision is computationally expensive, due to having to find corresponding pixels. Correlation is a fast, standard way to solve the correspondence problem. This paper analyses the behaviour of correlation based stereo to find ways to improve its quality while maintaining its realtime suitability. Three methods are suggested. Two of them aim to improve the disparity image especially at depth discontinuities, while one targets the identification of possible errors in general. Results are given on real stereo images with ground truth. A comparison with five standard correlation methods shows that improvements of simple stereo correlation are possible in real-time on current computer hardware.

Most visual features are parametric in nature, including, edges, lines, corners, and junctions. We propose an algorithm to automatically construct detectors for arbitrary parametric features. To maximize robustness we use realistic multi-parameter feature models and incorporate optical and sensing effects. Each feature is represented as a densely sampled parametric manifold in a low dimensional subspace of a Hilbert space. During detection, the vector of intensity values in a window about each pixel in the image is projected into the subspace. If the projection lies sufficiently close to the feature manifold, the feature is detected and the location of the closest manifold point yields the feature parameters. The concepts of parameter reduction by normalization, dimension reduction, pattern rejection, and heuristic search are all employed to achieve the required efficiency. Detectors have been constructed for five features, namely, step edge (five parameters), roof edge (five parameters), line (six parameters), corner (five parameters), and circular disc (six parameters). The results of detailed experiments are presented which demonstrate the robustness of feature detection and the accuracy of parameter estimation.

Given a collection of similar signals that have been deformed with respect to each other, the general signal-matching problem is to recover the deformation. We formulate the problem as the minimization of an energy measure that combines a smoothness term and a similarity term. The minimization reduces to a dynamic system governed by a set of coupled, first-order differential equations. The dynamic system finds an optimal solution at a coarse scale and then tracks it continuously to a fine scale. Among the major themes in recent work on visual signal matching have been the notions of matching as constrained opti-mization, of variational surface reconstruction, and of coarse-to-fine matching. Our solution captures these in a precise, succinct, and unified form. Results are presented for one-dimensional signals, a motion sequence, and a stereo pair. 1