Slow and Smooth: A Bayesian Theory for the Combination of Local Motion Signals in Human Vision

In order to estimate the motion of an object, the visual system needs to combine multiple local measurements, each of which carries some degree of ambiguity. We present a model of motion perception whereby measurements from different image regions are combined according to a Bayesian estimator --- the estimated motion maximizes the posterior probability assuming a prior favoring slow and smooth velocities. In reviewing a large number of previously published phenomena we find that the Bayesian estimator predicts a wide range of psychophysical results. This suggests that the seemingly complex set of illusions arise from a single computational strategy that is optimal under reasonable assumptions.

Curved Inertia Frames: Visual Attention and Perceptual Organization Using Convexity and Symmetry

In this paper we present an approach to perceptual organization and attention based on Curved Inertia Frames (C.I.F.), a novel definition of "curved axis of inertia'' tolerant to noisy and spurious data. The definition is useful because it can find frames that correspond to large, smooth, convex, symmetric and central parts. It is novel because it is global and can detect curved axes. We discuss briefly the relation to human perception, the recognition of non-rigid objects, shape description, and extensions to finding "features", inside/outside relations, and long- smooth ridges in arbitrary surfaces.

Fast Perceptual Learning in Visual Hyperacuity

In many different spatial discrimination tasks, such as in determining the sign of the offset in a vernier stimulus, the human visual system exhibits hyperacuity-level performance by evaluating spatial relations with the precision of a fraction of a photoreceptor"s diameter. We propose that this impressive performance depends in part on a fast learning process that uses relatively few examples and occurs at an early processing stage in the visual pathway. We show that this hypothesis is plausible by demonstrating that it is possible to synthesize, from a small number of examples of a given task, a simple (HyperBF) network that attains the required performance level. We then verify with psychophysical experiments some of the key predictions of our conjecture. In particular, we show that fast timulus-specific learning indeed takes place in the human visual system and that this learning does not transfer between two slightly different hyperacuity tasks.

Visual Tracking

A typical robot vision scenario might involve a vehicle moving with an unknown 3D motion (translation and rotation) while taking intensity images of an arbitrary environment. This paper describes the theory and implementation issues of tracking any desired point in the environment. This method is performed completely in software without any need to mechanically move the camera relative to the vehicle. This tracking technique is simple an inexpensive. Furthermore, it does not use either optical flow or feature correspondence. Instead, the spatio-temporal gradients of the input intensity images are used directly. The experimental results presented support the idea of tracking in software. The final result is a sequence of tracked images where the desired point is kept stationary in the images independent of the nature of the relative motion. Finally, the quality of these tracked images are examined using spatio-temporal gradient maps.

Pattern Motion Perception: Feature Tracking or Integration of Component Motions?

A key question regarding primate visual motion perception is whether the motion of 2D patterns is recovered by tracking distinctive localizable features [Lorenceau and Gorea, 1989; Rubin and Hochstein, 1992] or by integrating ambiguous local motion estimates [Adelson and Movshon, 1982; Wilson and Kim, 1992]. For a two-grating plaid pattern, this translates to either tracking the grating intersections or to appropriately combining the motion estimates for each grating. Since both component and feature information are simultaneously available in any plaid pattern made of contrast defined gratings, it is unclear how to determine which of the two schemes is actually used to recover the plaid"s motion. To address this problem, we have designed a plaid pattern made with subjective, rather than contrast defined, gratings. The distinguishing characteristic of such a plaid pattern is that it contains no contrast defined intersections that may be tracked. We find that notwithstanding the absence of such features, observers can accurately recover the pattern velocity. Additionally we show that the hypothesis of tracking "illusory features" to estimate pattern motion does not stand up to experimental test. These results present direct evidence in support of the idea that calls for the integration of component motions over the one that mandates tracking localized features to recover 2D pattern motion. The localized features, we suggest, are used primarily as providers of grouping information - which component motion signals to integrate and which not to.

Virtual Visual Hulls: Example-Based 3D Shape Estimation from a Single Silhouette

Recovering a volumetric model of a person, car, or other object of interest from a single snapshot would be useful for many computer graphics applications. 3D model estimation in general is hard, and currently requires active sensors, multiple views, or integration over time. For a known object class, however, 3D shape can be successfully inferred from a single snapshot. We present a method for generating a ``virtual visual hull''-- an estimate of the 3D shape of an object from a known class, given a single silhouette observed from an unknown viewpoint. For a given class, a large database of multi-view silhouette examples from calibrated, though possibly varied, camera rigs are collected. To infer a novel single view input silhouette's virtual visual hull, we search for 3D shapes in the database which are most consistent with the observed contour. The input is matched to component single views of the multi-view training examples. A set of viewpoint-aligned virtual views are generated from the visual hulls corresponding to these examples. The 3D shape estimate for the input is then found by interpolating between the contours of these aligned views. When the underlying shape is ambiguous given a single view silhouette, we produce multiple visual hull hypotheses; if a sequence of input images is available, a dynamic programming approach is applied to find the maximum likelihood path through the feasible hypotheses over time. We show results of our algorithm on real and synthetic images of people.

Sharing visual features for multiclass and multiview object detection

We consider the problem of detecting a large number of different classes of objects in cluttered scenes. Traditional approaches require applying a battery of different classifiers to the image, at multiple locations and scales. This can be slow and can require a lot of training data, since each classifier requires the computation of many different image features. In particular, for independently trained detectors, the (run-time) computational complexity, and the (training-time) sample complexity, scales linearly with the number of classes to be detected. It seems unlikely that such an approach will scale up to allow recognition of hundreds or thousands of objects. We present a multi-class boosting procedure (joint boosting) that reduces the computational and sample complexity, by finding common features that can be shared across the classes (and/or views). The detectors for each class are trained jointly, rather than independently. For a given performance level, the total number of features required, and therefore the computational cost, is observed to scale approximately logarithmically with the number of classes. The features selected jointly are closer to edges and generic features typical of many natural structures instead of finding specific object parts. Those generic features generalize better and reduce considerably the computational cost of an algorithm for multi-class object detection.

The Role of Fixation and Visual Attention in Object Recognition

This research project is a study of the role of fixation and visual attention in object recognition. In this project, we build an active vision system which can recognize a target object in a cluttered scene efficiently and reliably. Our system integrates visual cues like color and stereo to perform figure/ground separation, yielding candidate regions on which to focus attention. Within each image region, we use stereo to extract features that lie within a narrow disparity range about the fixation position. These selected features are then used as input to an alignment-style recognition system. We show that visual attention and fixation significantly reduce the complexity and the false identifications in model-based recognition using Alignment methods. We also demonstrate that stereo can be used effectively as a figure/ground separator without the need for accurate camera calibration.

Time-Frequency Representations for Speech Signals

This work addresses two related questions. The first question is what joint time-frequency energy representations are most appropriate for auditory signals, in particular, for speech signals in sonorant regions. The quadratic transforms of the signal are examined, a large class that includes, for example, the spectrograms and the Wigner distribution. Quasi-stationarity is not assumed, since this would neglect dynamic regions. A set of desired properties is proposed for the representation: (1) shift-invariance, (2) positivity, (3) superposition, (4) locality, and (5) smoothness. Several relations among these properties are proved: shift-invariance and positivity imply the transform is a superposition of spectrograms; positivity and superposition are equivalent conditions when the transform is real; positivity limits the simultaneous time and frequency resolution (locality) possible for the transform, defining an uncertainty relation for joint time-frequency energy representations; and locality and smoothness tradeoff by the 2-D generalization of the classical uncertainty relation. The transform that best meets these criteria is derived, which consists of two-dimensionally smoothed Wigner distributions with (possibly oriented) 2-D guassian kernels. These transforms are then related to time-frequency filtering, a method for estimating the time-varying 'transfer function' of the vocal tract, which is somewhat analogous to ceptstral filtering generalized to the time-varying case. Natural speech examples are provided. The second question addressed is how to obtain a rich, symbolic description of the phonetically relevant features in these time-frequency energy surfaces, the so-called schematic spectrogram. Time-frequency ridges, the 2-D analog of spectral peaks, are one feature that is proposed. If non-oriented kernels are used for the energy representation, then the ridge tops can be identified, with zero-crossings in the inner product of the gradient vector and the direction of greatest downward curvature. If oriented kernels are used, the method can be generalized to give better orientation selectivity (e.g., at intersecting ridges) at the cost of poorer time-frequency locality. Many speech examples are given showing the performance for some traditionally difficult cases: semi-vowels and glides, nasalized vowels, consonant-vowel transitions, female speech, and imperfect transmission channels.

Visual Navigation: Constructing and Utilizing Simple Maps of an Indoor Environment

The goal of this work is to navigate through an office environmentsusing only visual information gathered from four cameras placed onboard a mobile robot. The method is insensitive to physical changes within the room it is inspecting, such as moving objects. Forward and rotational motion vision are used to find doors and rooms, and these can be used to build topological maps. The map is built without the use of odometry or trajectory integration. The long term goal of the project described here is for the robot to build simple maps of its environment and to localize itself within this framework.

The Role of Knowledge in Visual Shape Representation

This report shows how knowledge about the visual world can be built into a shape representation in the form of a descriptive vocabulary making explicit the important geometrical relationships comprising objects' shapes. Two computational tools are offered: (1) Shapestokens are placed on a Scale-Space Blackboard, (2) Dimensionality-reduction captures deformation classes in configurations of tokens. Knowledge lies in the token types and deformation classes tailored to the constraints and regularities ofparticular shape worlds. A hierarchical shape vocabulary has been implemented supporting several later visual tasks in the two-dimensional shape domain of the dorsal fins of fishes.

Learning Shape Descriptions: Generating and Generalizing Models of Visual Objects

We present the results of an implemented system for learning structural prototypes from grey-scale images. We show how to divide an object into subparts and how to encode the properties of these subparts and the relations between them. We discuss the importance of hierarchy and grouping in representing objects and show how a notion of visual similarities can be embedded in the description language. Finally we exhibit a learning algorithm that forms class models from the descriptions produced and uses these models to recognize new members of the class.

Computer System for Visual Recognition Using Active Knowledge

A system for visual recognition is described, with implications for the general problem of representation of knowledge to assist control. The immediate objective is a computer system that will recognize objects in a visual scene, specifically hammers. The computer receives an array of light intensities from a device like a television camera. It is to locate and identify the hammer if one is present. The computer must produce from the numerical "sensory data" a symbolic description that constitutes its perception of the scene. Of primary concern is the control of the recognition process. Control decisions should be guided by the partial results obtained on the scene. If a hammer handle is observed this should suggest that the handle is part of a hammer and advise where to look for the hammer head. The particular knowledge that a handle has been found combines with general knowledge about hammers to influence the recognition process. This use of knowledge to direct control is denoted here by the term "active knowledge". A descriptive formalism is presented for visual knowledge which identifies the relationships relevant to the active use of the knowledge. A control structure is provided which can apply knowledge organized in this fashion actively to the processing of a given scene.

Computer Recognition of Three-Dimensional Objects in a Visual Scene

Methods are presented (1) to partition or decompose a visual scene into the bodies forming it; (2) to position these bodies in three-dimensional space, by combining two scenes that make a stereoscopic pair; (3) to find the regions or zones of a visual scene that belong to its background; (4) to carry out the isolation of objects in (1) when the input has inaccuracies. Running computer programs implement the methods, and many examples illustrate their behavior. The input is a two-dimensional line-drawing of the scene, assumed to contain three-dimensional bodies possessing flat faces (polyhedra); some of them may be partially occluded. Suggestions are made for extending the work to curved objects. Some comparisons are made with human visual perception. The main conclusion is that it is possible to separate a picture or scene into the constituent objects exclusively on the basis of monocular geometric properties (on the basis of pure form); in fact, successful methods are shown.