Search

Registration and modeling of shapes are two important problems in computer vision and pattern recognition. Despite enormous progress made over the past decade, these problems are still open. In this paper, we advance the state of the art in both directions. First, we consider an efficient registration method that aims to recover a one-to-one correspondence between shapes and introduce measures of uncertainties driven from the data which explain the local support of the recovered transformations. To this end, a free-form deformation is used to describe the deformation model. The transformation is combined with an objective function defined in the space of implicit functions used to represent shapes. Once the registration parameters have been recovered, we introduce a novel technique for model building and statistical interpretation of the training examples based on a variable bandwidth kernel approach. The support on the kernels varies spatially and is determined according to the uncertainties of the registration process. Such a technique introduces the ability to account for potential registration errors in the model. Handwritten character recognition and knowledge-based object extraction in medical images are examples of applications that demonstrate the potentials of the proposed framework. Index Terms--Shape registration, shape modeling, uncertainty estimates, nonparametric statistics, variable metric kernels.

A computer vision system is presented that can be used to acquire position data for an experiment where human observations have traditionally been used. A simple algorithm based on color discrimination is used to automatically locate an object of interest. Experiments can be monitored in real time using inexpensive cameras. Examples are given to illustrate the applicability of this system for autonomous data acquisition. [DOI: 10.1119/1.2348892]

A model that characterizes a one-color per spatial position image as a coding into luminance and chrominance of the corresponding three colors per spatial position image is defined. The model shows that visual artifacts after demosaicing are due aliasing between luminance and chrominance and could be solved using a preprocessing filter.

An algorithm for computing three-dimensional (3-D) velocity field and motion parameters from a range image sequence is presented. It is based on a new integral 3-D rigid motion invariant feature - the trace of a 3 x 3 "feature matrix" related to the moment of inertia tensor. This trace can be computed at every point of a moving surface and provides a quantitative measure of the local shape of the surface. Based on the feature's conservation along the trajectory of every moving point, a 3-D Flow Constraint Equation is formulated and solved for the velocity field. The robustness of the feature in presence of noise and discontinuity is analyzed. Index Terms - Motion analysis, optical flow, rigid motion invariants, stereo, 3-D flow, 3-D motion detection.

Conventional computer vision methods for determining a robot's end-effector motion based on sensory data needs sensor calibration (e.g., camera calibration) and sensor-to-hand calibration (e.g., hand-eye calibration). This involves many computations and even some difficulties, especially when different kinds of sensors are involved. In this correspondence, we present a neural network approach to the motion determination problem without any calibration. Two kinds of sensory data, namely, camera images and laser range data, are used as the input to a multilayer feedforward network to associate the direct transformation from the sensory data to the required motions. This provides a practical sensor fusion method. Using a recursive motion strategy and in terms of a network correction, we relax the requirement for the exactness of the learned transformation. Another important feature of our work is that the goal position can be changed without having to do network retraining. Experimental results show the effectiveness of our method. Index Terms - Cameras, laser range finders, neural networks, visual servoing.

Standard least-squares (LS) methods for pose estimation of objects are sensitive to outliers which can occur due to mismatches. Even a single mismatch can severely distort the estimated pose. This paper describes a least-median of squares (LMedS) approach to estimating pose using point matches. It is both robust (resistant to up to 50% outliers) and efficient (linear in the number of points). The basic algorithm is then extended to improve performance in the presence of two types of noise: 1) type I which perturbs all data values by small amounts (e.g., Gaussian) and 2) type II which can corrupt a few data values by large amounts. Index Terms - Least median squares, robust pose estimation.

We investigate the exploitation of linear models of camera-object interaction for an efficient modeling and control of image-based visual servoing systems. The approach includes a method for coping with those representation ambiguities typical of linear interaction models which may affect both planning and control. The implementation of an eye-in-hand servoing system based on affine camera models and using image contours as relevant visual features is described and discussed. The system, including an image planner, a two-dimensional/three-dimensional (2-D/3-D) controller, and a visual analysis module, allows an intuitive specification and execution of relative positioning tasks w.r.t. still or moving rigid objects. Results of real-time experiments with a robotic platform featuring a PUMA manipulator provide a further insight into characteristics and performance of affine visual servoing systems. Index Terms - Affine interaction models, eye-in-hand systems, visual servoing.

This paper introduces a new algorithm called Lane-finding in ANother domAin (LANA) for detecting lane markers in images acquired from a forward-looking vehicle-mounted camera. The method is based on a novel set of frequency domain features that capture relevant information concerning the strength and orientation of spatial edges. The frequency domain features are combined with a deformable template prior, in order to detect the lane markers of interest. Experimental results that illustrate the performance of this algorithm on images with varying lighting and environmental conditions, shadowing, lane occlusion(s), solid and dashed lines, etc. are presented. LANA seems to detect lane markers remarkably well under a very large and varied collection of roadway images. A comparison is drawn between this frequency feature-based LANA algorithm and the spatial feature-based LOIS lane detection algorithm. This comparison is made from experimental, computational, and methodological standpoints. Index Terms - Bayesian estimation, denoising, discrete cosine transform, global shape model, intelligent vehicles, likelihood ratio, number of operations.

This paper presents a new intensity-to-time processing paradigm suitable for very large scale integration (VLSI) computational sensor implementation of global operations over sensed images. Global image quantities usually describe images with fewer data. When computed at the point of sensing, global quantities result in a low-latency performance due to the reduced data transfer requirements between an image sensor and a processor. The global quantities also help global top-down adaptation: the quantities are continuously computed on-chip, and are readily available to sensing for adaptation. As an example, we have developed a sorting image computational sensor - a VLSI chip which senses an image and sorts all pixel by their intensities. The first sorting sensor prototype is a 21 x 26 array of cells. It receives an image optically, senses it, and computes the image's cumulative histogram - a global quantity which can be quickly routed off chip via one pin. In addition, the global cumulative histogram is used internally on-chip in a top-down fashion to adapt the values in individual pixel so as to reflect the index of the incoming light, thus computing an "image of indices." The image of indices never saturates and has a uniform histogram. Index Terms - Computational sensors, image sensors, robot perception, smart sensors, vision, VLSI sensors.

In this paper, we present the two-dimensional (2-D) version of the symmetries and perturbation model (SPmodel), a probabilistic representation model and an EKF integration mechanism for uncertain geometric information that is suitable for sensor fusion and integration in multisensor systems. We apply the SPmodel to the problem of location estimation in indoor mobile robotics, experimenting with the mobile robot MACROBE. We have chosen two types of complementary sensory information: 1) range images; 2) intensity images; obtained from a laser sensor. Results of these experiments show that fusing simple and computationally inexpensive sensory information can allow a mobile robot to precisely locate itself. They also demonstrate the generality of the proposed fusion and integration mechanism. Index Terms - Computer vision, map-based localization, mobile robots, range finder, sensor fusion.

An analytically tractable potential field model of free space is presented in this paper. The model assumes that the border of every two-dimensional (2-D) region is uniformly charged. It is shown that the potential and the resulting repulsion (force and torque) between polygonal regions can be calculated in closed form. By using the Newtonian potential function, collision avoidance between object and obstacle thus modeled is guaranteed in a path planning problem. A local planner is developed for finding object paths going through narrow areas of free space where the obstacle avoidance is most important. Simulation results show that not only does individual object configuration of a path obtained with the proposed approach avoid obstacles effectively, the configurations also connect smoothly into a path. Index Terms - Artificial potential field, free-space representation, obstacle avoidance, path planning.

It is well-known that the linear scale-space theory in computer vision is mainly based on the Gaussian kernel. The purpose of the paper is to propose a scale-space theory based on B-spline kernels. Our aim is twofold. On one hand, we present a general framework and show how B-splines provide a flexible tool to design various scale-space representations: continuous scale-space, dyadic scale-space frame, and compact scale-space representation. In particular, we focus on the design of continuous scale-space and dyadic scale-space frame representation. A general algorithm is presented for fast implementation of continuous scale-space at rational scales. In the dyadic case, efficient frame algorithms are derived using B-spline techniques to analyze the geometry of an image. Moreover, the image can be synthesized from its multiscale local partial derivatives. Also, the relationship between several scale-space approaches is explored. In particular, the evolution of wavelet theory from traditional scale-space filtering can be well understood in terms of B-splines. On the other hand, the behavior of edge models, the properties of completeness, causality, and other properties in such a scale-space representation are examined in the framework of B-splines. It is shown that, besides the good properties inherited from the Gaussian kernel, the B-spline derived scale-space exhibits many advantages for modeling visual mechanism with regard to the efficiency, compactness, orientation feature, and parallel structure. Index Terms - Image modeling, B-spline, wavelet, scale-space, scaling theorem, fingerprint theorem.

The literature on recursive estimation of structure and motion from monocular image sequences comprises a large number of apparently unrelated models and estimation techniques. We propose a framework that allows us to derive and compare all models by following the idea of dynamical system reduction. The "natural" dynamic model, derived from the rigidity constraint and the projection model, is first reduced by explicitly decoupling structure (depth) from motion. Then, implicit decoupling techniques are explored, which consist of imposing that some function of the unknown parameters is held constant. By appropriately choosing such a function. not only can we account for models seen so far in the literature, but we can also derive novel ones. Index Terms - Visual motion estimation, epipolar geometry, motion decoupling, compensation, fixation, parallax, output stabilization, model reduction.

This paper concerns the extraction of rotation invariant texture features and the use of such features in script identification from document images. Rotation invariant texture features are computed based on an extension of the popular multi-channel Gabor filtering technique, and their effectiveness is tested with 300 randomly rotated samples of 15 Brodatz textures. These features are then used in an attempt to solve a practical but hitherto mostly overlooked problem in document image processing - the identification of the script of a machine printed document. Automatic script and language recognition is an essential front-end process for the efficient and correct use of OCR and language translation products in a multilingual environment. Six languages (Chinese, English, Greek, Russian, Persian, and Malayalam) are chosen to demonstrate the potential of such a texture-based approach in script identification. Index Terms - Texture, rotation invariance, Gabor filtering, script and language recognition, document image processing.

The three best-known criteria in two-view motion analysis are based, respectively, on the distances between points and their corresponding epipolar lines, on the gradient-weighted epipolar errors, and on the distances between points and the reprojections of their reconstructed points. The last one has a better statistical interpretation, but is, however, significantly slower than the first two. In this paper, I show that, given a reasonable initial guess of the epipolar geometry, the last two criteria are equivalent when the epipoles are at infinity, and differ from each other only a little even when the epipoles are in the image, as shown experimentally. The first two criteria are equivalent only when the epipoles are at infinity and when the observed object/scene has the same scale in the two images. This suggests that the second criterion is sufficient in practice because of its computational efficiency. Experiments with several thousand computer simulations and four sets of real data confirm the analysis. The result is valid for both calibrated and uncalibrated images. Index Terms - Motion analysis, multiple-view geometry, 3D reconstruction, optimization criteria, algorithmic comparison, structure from motion, uncalibrated images.

Junctions are important features for image analysis and form a critical aspect of image understanding tasks such as object recognition. We present a unified approach to detecting (location of the center of the junction), classifying (by the number of wedges - lines, corners, three-junctions such as T or Y junctions, or four-junctions such as X-junctions), and reconstructing junctions (in terms of radius size, the angles of each wedge and the intensity in each of the wedges) in images. Our main contribution is a modeling of the junction which is complex enough to handle all these issues and yet simple enough to admit an effective dynamic programming solution. Broadly, we use a template deformation framework along with a gradient criterium to detect radial partitions of the template. We use the minimum description length principle to obtain the optimal number of partitions that best describes the junction. Kona [27] is an implementation of this model. We (quantitatively) demonstrate the stability and robustness of the detector by analyzing its behavior in the presence of noise, using synthetic/controlled apparatus. We also present a qualitative study of its behavior on real images. Index Terms - Junctions, corners, feature detection, low-level vision, minimum description length (MDL) principle, energy minimization.

This paper considers the problem of systematically exploring an unfamiliar environment in search of one or more recognizable targets. The proposed exploration algorithm is based on a novel representation of environments containing visual landmarks called the boundary place graph. This representation records the set of recognizable objects (landmarks) that are visible from the boundary of each configuration space obstacle. No metric information about the scene geometry is recorded nor are explicit prescriptions for moving between places stored. The exploration algorithm constructs the boundary place graph incrementally from sensor data. Once the robot has completely explored an environment, it can use the constructed representation to carry out further navigation tasks. In order to precisely characterize the set of environments in which this algorithm is expected to succeed, we provide a necessary and sufficient condition under which the algorithm is guaranteed to discover all landmarks. This algorithm has been implemented on our mobile robot platform R J, and results from these experiments are presented. Importantly, this research demonstrates that it is possible to design and implement provably correct exploration and navigation algorithms that do not require global positioning systems or metric representations of the environment. Index Terms - Exploration, landmarks, mobile robots, navigation.

Despite the intuitive interpretation of the skeleton as the "center line of the shape," it is surprisingly hard to get skeletonization algorithms that simultaneously produce all the "good" properties we expect (e.g., well-centered, well-connected, rotation-invariant, efficient, robust, accurately reflecting the shape). In this paper, we introduce the new concept of "veinerization," which produces a graph that contains all the "topological" information needed to derive a wide variety of skeletons. Theoretically, the main contribution is to provide a homogeneous framework for integration of the major concepts described in other related works on digital skeletonization. In practice, the new aspect of this approach is to provide the user with different criteria for selecting the most suitable skeleton for a given application; e.g., the user can select a suitable threshold for obtaining the desirable balance between "having a skeleton without noisy prunes" and "having a skeleton that reflects the initial shape." This algorithm has been tested on numerous kinds of patterns, including pathological ones like fractal sets well-known for the complexity of their shapes. Index Terms - Digital skeleton, veinerization, skeletonization, context adaptation, digital topology, graphs, fractal sets.

A geometrical framework for low-level vision which uses borrowed ideas from high-energy physics is proposed. The framework was developed to define images as embedding maps between two Riemannian manifolds and to measure the space of the maps. The framework's use in computer vision is justified by it's ability to merge many seemingly-unrelated scale space processes, while introducing innovative ways to denoise and smooth images. In addition, the framework integrates several existing scale space and denoising techniques through changing a single parameter which switches between the Euclidean L1 and L2 norms.

This paper presents a systematic profile recognition and mensuration approach in machine vision. It can be utilized to recognize and measure the profiles of industrial parts in an automated manufacturing process by machine vision systems. A new method of profile representation by sampling the data from the object boundary in a digital image is presented. Autoregressive (AR) models are used to code the sampled data of the profiles into AR coefficients for profile recognition. Characterization of the profiles is accomplished by the Data Dependent Systems (DDS) methodology. The AR coefficients and characteristic roots help construct the AR and DDS descriptors to characterize the signatures of the profiles. The frequency domain information about the profiles can be extracted by DDS analysis. The measurement of the profile variation is obtained from the DDS results using optical mensuration method. Neural network and feature weighting method are utilized as reasoning machines for recognition. The illustrative examples in which the profile sampled data are corrupted by noise show that the profile recognition and mensuration approach is very effective and robust in a typical noisy environment on the shop floor.

A navigation system is developed to show how a dynamic approach can be used in representing spatial information and integrating multiple information sources and behaviors. The system features a vehicle which estimates ego-position based on either dead-reckoning or the recognition of local views associated with memorized positions. It is shown that the system does not have an absolute Cartesian map of home base position and orientations but rather recalibrates from one error prone local reference frame to the next.

This paper presents the results of experiments performed to validate the computer retina model presented in a companion paper [15]. Experiments commonly performed by electrophysiologists on biological retinas are simulated and the computer retina outputs compared with published recordings of cells in biological (principally primate) retinas. Experiments with more complex stimuli further reveal how the computer retina enhances spatio-temporal contrast information and adapts to a wide range of illumination levels much like the primate retina.

Many vision research groups have developed the active vision platform whereby the camera motion can be controlled. A similar setup is the wrist-mounted camera for a robot manipulator. This head-eye (or hand-eye) setup considerably facilitates motion stereo, object tracking, and active perception. One of the important issues in using the active vision system is to determine the camera position and orientation relative to the camera platform. This problem is called the head-eye calibration in active vision, and the hand-eye calibration in robotics. In this paper we present a new technique for calibrating the head-eye (or hand-eye) geometry as well as the camera intrinsic parameters. The technique allows camera self-calibration because it requires no reference object and directly uses the images of the environment. Camera self-calibration is important especially in circumstances where the execution of the underlying visual tasks does not permit the use of reference objects. Our method exploits the flexibility of the active vision system, and bases camera calibration on a sequence of specially designed motion. It is shown that if the camera intrinsic parameters are known a priori, the orientation of the camera relative to the platform can be solved using 3 pure translational motions. If the intrinsic parameters are unknown, then two sequences of motion, each consisting of three orthogonal translations, are necessary to determine the camera orientation and intrinsic parameters. Once the camera orientation and intrinsic parameters are determined, the position of the camera relative to the platform can be computed from an arbitrary nontranslational motion of the platform. All the computations in our method are linear. Experimental results with real images are presented in this paper.

The visual search process is modeled as a sequence of fixations which may or may not partially overlap in coverage of the search field. A single expression is derived for the probability of search success which covers all overlap cases. Making the further assumption that search strategy is chosen so as to maximize expected value, yields the result that each point on the search field must be fixated an integral number of times. In an experiment where visual lobe size and fixation duration were controlled, human subjects behaved in a manner close to that predicted by the optimization model. It is concluded that further development of the model using less restrictive assumptions is warranted.

A variety of analytic and probabilistic models in connection to Markov random fields (MRFs) have been proposed in the last decade for solving low level vision problems involving discontinuities. This paper presents a systematic study of these models and defines a general discontinuity adaptive (DA) MRF model. By analyzing the Euler equation associated with the energy minimization, it shows that the fundamental difference between different models lies in the behavior of interaction between neighboring points, which is determined by the a priori smoothness constraint encoded into the energy function. An important necessary condition is derived for the interaction to be adaptive to discontinuities to avoid oversmoothing. This forms the basis on which a class of adaptive interaction functions (AIFs) is defined. The DA model is defined in terms of the Euler equation constrained by this class of AIFs. Its solution is [C.sup.1] continuous and allows arbitrarily large but bounded slopes in dealing with dis-continuities. Because of the continuous nature, it is stable to changes in parameters and data, a good property for regularizing ill-posed problems. Experimental results are shown. Index Terms - Discontinuities, energy functions, Euler equation, computer vision, Markov random fields, minimization, regularization.

This paper considers the problem of navigating a point robot in an unknown two-dimensional terrain populated by disjoint generalized polygonal obstacles. A generalized polygon consists of a connected sequence of circular arcs and straight-line segments. The terrain model is not known a priori, but the robot is equipped with a vision sensor. A discrete vision sensor detects all visible (from a single position) portions of the obstacle boundaries in a single scan operation. The navigation problem deals with moving the robot through the terrain from a source position to a destination position, and the terrain model acquisition problem deals with autonomously building a model of the terrain. A complete solution to either problem is shown to require an infinite number of scan operations in cusp regions formed by a pair of convex and concave obstacle edges. Either problem is considered solved with a precision [Epsilon] if the points that have not been scanned are those in a cusp region with a clearance less than [Epsilon] from two obstacle edges. Three methods are proposed to solve both problems with a precision [Epsilon] based on extensions of the generalized visibility graph, the generalized Voronoi diagram, and the trapezoidal decomposition. Then simplified versions of these structures are proposed to exactly solve the navigation and terrain model acquisition problems using a continuous vision sensor that detects all visible obstacle boundaries as the robot navigates along a path.

This paper presents a system for modeling solids of revolution from a set of images taken with a zoom-lens. Using a zoom in monocular vision has interesting optical properties. The displacement of the optical center along the optical axis, when the focal length is modified, permits to implement a triangulation process. Modeling a solid of revolution requires the estimation of the 3D location of its revolution axis. First, we describe an original method, using zoom properties, to compute the 3D pose of the revolution axis. It requires the detection in each image of the object axis projection. Then, the modeling algorithm is presented. It is based on the resolution of the inverse perspective problem for points detected in the images. An experimental result of reconstruction, from real images set is finally given.

Gibbs random field (GRF) models and features from cooccurrence matrices are typically considered as separate but useful tools for texture discrimination. In this paper we show an explicit relationship between cooccurrences and a large class of GRF's. This result comes from a new framework based on a set-theoretic concept called the 'aura set' and on measures of this set, 'aura measures.' This framework is also shown to be useful for relating different texture analysis tools: We show how the aura set can be constructed with morphological dilation, how its measure yields cooccurrences, and how it can be applied to characterizing the behavior of the Gibbs model for texture. In particular, we show how the aura measure generalizes, to any number of gray levels and neighborhood order, some properties previously known for just the binary, nearest-neighbor GRF. Finally, we illustrate how these properties can guide one's intuition about the types of GRF patterns which are most likely to form.

An image representation scheme using a set of block pattern models (BPM's) is introduced. These models consist of three categories (constant, oriented, and irregular), with the last one consisting of two subgroups: textured and mixed. They are constructed to represent three basic types of image patterns: shade, edge, and texture. Image representation using these models requires considerably fewer bits than the original pixel-wise description and yet characterizes perceptually significant features more effectively. In particular, the parameterization of the oriented model using oriented basis functions lays a mathematical foundation for designing directional operators that are desirable in many image processing applications. Algorithms for model classification, model parameter estimation, and image reconstruction from model parameters are presented, and these provide the necessary vehicles for applying the proposed representation scheme to various image processing tasks. The applications of the proposed models in image coding, image zooming, and image smoothing are described. The coding system approximates each image block by a BPM and further quantizes the model parameters. Satisfactory coded images have been obtained at bit rates between 0.5 approximates 0.6 bpp (bits per pixel) with a high-rate realization and between 0.3 approximates 0.5 bpp with a low-rate realization. The high-rate realization has a simple structure suitable for real-time implementation. The methods for image zooming and smoothing are similar, where both adapt the processing for each pixel according to the model of its neighborhood. By using directional filters in oriented regions, edges and lines are rendered sharper in a smoother manner than with conventional linear filtering approaches, which leads to significant improvement in perceived image quality.

Although most optical flow techniques presume brightness constancy, it is well-known that this constraint is often violated, producing poor estimates of image motion. This paper describes a generalized formulation of optical flow estimation based on models of brightness variations that are caused by time-dependent physical processes. These include changing surface orientation with respect to a directional illuminant, motion of the illuminant, and physical models of heat transport in infrared images. With these models, we simultaneously estimate the 2D image motion and the relevant physical parameters of the brightness change model. The estimation problem is formulated using total least squares (TLS), with confidence bounds on the parameters. Experiments in four domains, with both synthetic and natural inputs, show how this formulation produces superior estimates of the 2D image motion. Index Terms--Optical flow, physics-based brightness variation, total least squares.