Probabilistic Solution of Inverse Problems

In this thesis we study the general problem of reconstructing a function, defined on a finite lattice from a set of incomplete, noisy and/or ambiguous observations. The goal of this work is to demonstrate the generality and practical value of a probabilistic (in particular, Bayesian) approach to this problem, particularly in the context of Computer Vision. In this approach, the prior knowledge about the solution is expressed in the form of a Gibbsian probability distribution on the space of all possible functions, so that the reconstruction task is formulated as an estimation problem. Our main contributions are the following: (1) We introduce the use of specific error criteria for the design of the optimal Bayesian estimators for several classes of problems, and propose a general (Monte Carlo) procedure for approximating them. This new approach leads to a substantial improvement over the existing schemes, both regarding the quality of the results (particularly for low signal to noise ratios) and the computational efficiency. (2) We apply the Bayesian appraoch to the solution of several problems, some of which are formulated and solved in these terms for the first time. Specifically, these applications are: teh reconstruction of piecewise constant surfaces from sparse and noisy observationsl; the reconstruction of depth from stereoscopic pairs of images and the formation of perceptual clusters. (3) For each one of these applications, we develop fast, deterministic algorithms that approximate the optimal estimators, and illustrate their performance on both synthetic and real data. (4) We propose a new method, based on the analysis of the residual process, for estimating the parameters of the probabilistic models directly from the noisy observations. This scheme leads to an algorithm, which has no free parameters, for the restoration of piecewise uniform images. (5) We analyze the implementation of the algorithms that we develop in non-conventional hardware, such as massively parallel digital machines, and analog and hybrid networks.

Exploiting Phonological Constraints for Handshape Inference in ASL Video

Handshape is a key articulatory parameter in sign language, and thus handshape recognition from signing video is essential for sign recognition and retrieval. Handshape transitions within monomorphemic lexical signs (the largest class of signs in signed languages) are governed by phonological rules. For example, such transitions normally involve either closing or opening of the hand (i.e., to exclusively use either folding or unfolding of the palm and one or more fingers). Furthermore, akin to allophonic variations in spoken languages, both inter- and intra- signer variations in the production of specific handshapes are observed. We propose a Bayesian network formulation to exploit handshape co-occurrence constraints, also utilizing information about allophonic variations to aid in handshape recognition. We propose a fast non-rigid image alignment method to gain improved robustness to handshape appearance variations during computation of observation likelihoods in the Bayesian network. We evaluate our handshape recognition approach on a large dataset of monomorphemic lexical signs. We demonstrate that leveraging linguistic constraints on handshapes results in improved handshape recognition accuracy. As part of the overall project, we are collecting and preparing for dissemination a large corpus (three thousand signs from three native signers) of American Sign Language (ASL) video. The video have been annotated using SignStream® [Neidle et al.] with labels for linguistic information such as glosses, morphological properties and variations, and start/end handshapes associated with each ASL sign.

The Specialized Mappings Architecture

A probabilistic, nonlinear supervised learning model is proposed: the Specialized Mappings Architecture (SMA). The SMA employs a set of several forward mapping functions that are estimated automatically from training data. Each specialized function maps certain domains of the input space (e.g., image features) onto the output space (e.g., articulated body parameters). The SMA can model ambiguous, one-to-many mappings that may yield multiple valid output hypotheses. Once learned, the mapping functions generate a set of output hypotheses for a given input via a statistical inference procedure. The SMA inference procedure incorporates an inverse mapping or feedback function in evaluating the likelihood of each of the hypothesis. Possible feedback functions include computer graphics rendering routines that can generate images for given hypotheses. The SMA employs a variant of the Expectation-Maximization algorithm for simultaneous learning of the specialized domains along with the mapping functions, and approximate strategies for inference. The framework is demonstrated in a computer vision system that can estimate the articulated pose parameters of a human’s body or hands, given silhouettes from a single image. The accuracy and stability of the SMA are also tested using synthetic images of human bodies and hands, where ground truth is known.

BoostMap: A Method for Efficient Approximate Similarity Rankings

This paper introduces BoostMap, a method that can significantly reduce retrieval time in image and video database systems that employ computationally expensive distance measures, metric or non-metric. Database and query objects are embedded into a Euclidean space, in which similarities can be rapidly measured using a weighted Manhattan distance. Embedding construction is formulated as a machine learning task, where AdaBoost is used to combine many simple, 1D embeddings into a multidimensional embedding that preserves a significant amount of the proximity structure in the original space. Performance is evaluated in a hand pose estimation system, and a dynamic gesture recognition system, where the proposed method is used to retrieve approximate nearest neighbors under expensive image and video similarity measures. In both systems, BoostMap significantly increases efficiency, with minimal losses in accuracy. Moreover, the experiments indicate that BoostMap compares favorably with existing embedding methods that have been employed in computer vision and database applications, i.e., FastMap and Bourgain embeddings.

Automated Placement of Cameras in a Floorplan to Satisfy Task-Specific Constraints

In many multi-camera vision systems the effect of camera locations on the task-specific quality of service is ignored. Researchers in Computational Geometry have proposed elegant solutions for some sensor location problem classes. Unfortunately, these solutions utilize unrealistic assumptions about the cameras' capabilities that make these algorithms unsuitable for many real-world computer vision applications: unlimited field of view, infinite depth of field, and/or infinite servo precision and speed. In this paper, the general camera placement problem is first defined with assumptions that are more consistent with the capabilities of real-world cameras. The region to be observed by cameras may be volumetric, static or dynamic, and may include holes that are caused, for instance, by columns or furniture in a room that can occlude potential camera views. A subclass of this general problem can be formulated in terms of planar regions that are typical of building floorplans. Given a floorplan to be observed, the problem is then to efficiently compute a camera layout such that certain task-specific constraints are met. A solution to this problem is obtained via binary optimization over a discrete problem space. In preliminary experiments the performance of the resulting system is demonstrated with different real floorplans.

Automated Camera Layout to Satisfy Task-Specific and Floorplan-Specific Coverage Requirements

In many multi-camera vision systems the effect of camera locations on the task-specific quality of service is ignored. Researchers in Computational Geometry have proposed elegant solutions for some sensor location problem classes. Unfortunately, these solutions utilize unrealistic assumptions about the cameras' capabilities that make these algorithms unsuitable for many real-world computer vision applications: unlimited field of view, infinite depth of field, and/or infinite servo precision and speed. In this paper, the general camera placement problem is first defined with assumptions that are more consistent with the capabilities of real-world cameras. The region to be observed by cameras may be volumetric, static or dynamic, and may include holes that are caused, for instance, by columns or furniture in a room that can occlude potential camera views. A subclass of this general problem can be formulated in terms of planar regions that are typical of building floorplans. Given a floorplan to be observed, the problem is then to efficiently compute a camera layout such that certain task-specific constraints are met. A solution to this problem is obtained via binary optimization over a discrete problem space. In experiments the performance of the resulting system is demonstrated with different real floorplans.

Layered Graphical Models for Tracking Partially-Occluded Objects

Partial occlusions are commonplace in a variety of real world computer vision applications: surveillance, intelligent environments, assistive robotics, autonomous navigation, etc. While occlusion handling methods have been proposed, most methods tend to break down when confronted with numerous occluders in a scene. In this paper, a layered image-plane representation for tracking people through substantial occlusions is proposed. An image-plane representation of motion around an object is associated with a pre-computed graphical model, which can be instantiated efficiently during online tracking. A global state and observation space is obtained by linking transitions between layers. A Reversible Jump Markov Chain Monte Carlo approach is used to infer the number of people and track them online. The method outperforms two state-of-the-art methods for tracking over extended occlusions, given videos of a parking lot with numerous vehicles and a laboratory with many desks and workstations.

Learning a Family of Detectors

Object detection and recognition are important problems in computer vision. The challenges of these problems come from the presence of noise, background clutter, large within class variations of the object class and limited training data. In addition, the computational complexity in the recognition process is also a concern in practice. In this thesis, we propose one approach to handle the problem of detecting an object class that exhibits large within-class variations, and a second approach to speed up the classification processes. In the first approach, we show that foreground-background classification (detection) and within-class classification of the foreground class (pose estimation) can be jointly solved with using a multiplicative form of two kernel functions. One kernel measures similarity for foreground-background classification. The other kernel accounts for latent factors that control within-class variation and implicitly enables feature sharing among foreground training samples. For applications where explicit parameterization of the within-class states is unavailable, a nonparametric formulation of the kernel can be constructed with a proper foreground distance/similarity measure. Detector training is accomplished via standard Support Vector Machine learning. The resulting detectors are tuned to specific variations in the foreground class. They also serve to evaluate hypotheses of the foreground state. When the image masks for foreground objects are provided in training, the detectors can also produce object segmentation. Methods for generating a representative sample set of detectors are proposed that can enable efficient detection and tracking. In addition, because individual detectors verify hypotheses of foreground state, they can also be incorporated in a tracking-by-detection frame work to recover foreground state in image sequences. To run the detectors efficiently at the online stage, an input-sensitive speedup strategy is proposed to select the most relevant detectors quickly. The proposed approach is tested on data sets of human hands, vehicles and human faces. On all data sets, the proposed approach achieves improved detection accuracy over the best competing approaches. In the second part of the thesis, we formulate a filter-and-refine scheme to speed up recognition processes. The binary outputs of the weak classifiers in a boosted detector are used to identify a small number of candidate foreground state hypotheses quickly via Hamming distance or weighted Hamming distance. The approach is evaluated in three applications: face recognition on the face recognition grand challenge version 2 data set, hand shape detection and parameter estimation on a hand data set, and vehicle detection and estimation of the view angle on a multi-pose vehicle data set. On all data sets, our approach is at least five times faster than simply evaluating all foreground state hypotheses with virtually no loss in classification accuracy.

Recognition of Human Action Using Moment-Based Feature

The performance of different classification approaches is evaluated using a view-based approach for motion representation. The view-based approach uses computer vision and image processing techniques to register and process the video sequence. Two motion representations called Motion Energy Images and Motion History Image are then constructed. These representations collapse the temporal component in a way that no explicit temporal analysis or sequence matching is needed. Statistical descriptions are then computed using moment-based features and dimensionality reduction techniques. For these tests, we used 7 Hu moments, which are invariant to scale and translation. Principal Components Analysis is used to reduce the dimensionality of this representation. The system is trained using different subjects performing a set of examples of every action to be recognized. Given these samples, K-nearest neighbor, Gaussian, and Gaussian mixture classifiers are used to recognize new actions. Experiments are conducted using instances of eight human actions (i.e., eight classes) performed by seven different subjects. Comparisons in the performance among these classifiers under different conditions are analyzed and reported. Our main goals are to test this dimensionality-reduced representation of actions, and more importantly to use this representation to compare the advantages of different classification approaches in this recognition task.

Detecting Instances of Shape Classes That Exhibit Variable Structure

This paper proposes a method for detecting shapes of variable structure in images with clutter. The term "variable structure" means that some shape parts can be repeated an arbitrary number of times, some parts can be optional, and some parts can have several alternative appearances. The particular variation of the shape structure that occurs in a given image is not known a priori. Existing computer vision methods, including deformable model methods, were not designed to detect shapes of variable structure; they may only be used to detect shapes that can be decomposed into a fixed, a priori known, number of parts. The proposed method can handle both variations in shape structure and variations in the appearance of individual shape parts. A new class of shape models is introduced, called Hidden State Shape Models, that can naturally represent shapes of variable structure. A detection algorithm is described that finds instances of such shapes in images with large amounts of clutter by finding globally optimal correspondences between image features and shape models. Experiments with real images demonstrate that our method can localize plant branches that consist of an a priori unknown number of leaves and can detect hands more accurately than a hand detector based on the chamfer distance.

Music Maker – A Camera-based Music Making Tool for Physical Rehabilitation

The therapeutic effects of playing music are being recognized increasingly in the field of rehabilitation medicine. People with physical disabilities, however, often do not have the motor dexterity needed to play an instrument. We developed a camera-based human-computer interface called "Music Maker" to provide such people with a means to make music by performing therapeutic exercises. Music Maker uses computer vision techniques to convert the movements of a patient's body part, for example, a finger, hand, or foot, into musical and visual feedback using the open software platform EyesWeb. It can be adjusted to a patient's particular therapeutic needs and provides quantitative tools for monitoring the recovery process and assessing therapeutic outcomes. We tested the potential of Music Maker as a rehabilitation tool with six subjects who responded to or created music in various movement exercises. In these proof-of-concept experiments, Music Maker has performed reliably and shown its promise as a therapeutic device.

Learning Embeddings for Indexing, Retrieval, and Classification, with Applications to Object and Shape Recognition in Image Databases

Nearest neighbor retrieval is the task of identifying, given a database of objects and a query object, the objects in the database that are the most similar to the query. Retrieving nearest neighbors is a necessary component of many practical applications, in fields as diverse as computer vision, pattern recognition, multimedia databases, bioinformatics, and computer networks. At the same time, finding nearest neighbors accurately and efficiently can be challenging, especially when the database contains a large number of objects, and when the underlying distance measure is computationally expensive. This thesis proposes new methods for improving the efficiency and accuracy of nearest neighbor retrieval and classification in spaces with computationally expensive distance measures. The proposed methods are domain-independent, and can be applied in arbitrary spaces, including non-Euclidean and non-metric spaces. In this thesis particular emphasis is given to computer vision applications related to object and shape recognition, where expensive non-Euclidean distance measures are often needed to achieve high accuracy. The first contribution of this thesis is the BoostMap algorithm for embedding arbitrary spaces into a vector space with a computationally efficient distance measure. Using this approach, an approximate set of nearest neighbors can be retrieved efficiently - often orders of magnitude faster than retrieval using the exact distance measure in the original space. The BoostMap algorithm has two key distinguishing features with respect to existing embedding methods. First, embedding construction explicitly maximizes the amount of nearest neighbor information preserved by the embedding. Second, embedding construction is treated as a machine learning problem, in contrast to existing methods that are based on geometric considerations. The second contribution is a method for constructing query-sensitive distance measures for the purposes of nearest neighbor retrieval and classification. In high-dimensional spaces, query-sensitive distance measures allow for automatic selection of the dimensions that are the most informative for each specific query object. It is shown theoretically and experimentally that query-sensitivity increases the modeling power of embeddings, allowing embeddings to capture a larger amount of the nearest neighbor structure of the original space. The third contribution is a method for speeding up nearest neighbor classification by combining multiple embedding-based nearest neighbor classifiers in a cascade. In a cascade, computationally efficient classifiers are used to quickly classify easy cases, and classifiers that are more computationally expensive and also more accurate are only applied to objects that are harder to classify. An interesting property of the proposed cascade method is that, under certain conditions, classification time actually decreases as the size of the database increases, a behavior that is in stark contrast to the behavior of typical nearest neighbor classification systems. The proposed methods are evaluated experimentally in several different applications: hand shape recognition, off-line character recognition, online character recognition, and efficient retrieval of time series. In all datasets, the proposed methods lead to significant improvements in accuracy and efficiency compared to existing state-of-the-art methods. In some datasets, the general-purpose methods introduced in this thesis even outperform domain-specific methods that have been custom-designed for such datasets.

Spatiotemporal Gesture Segmentation

Spotting patterns of interest in an input signal is a very useful task in many different fields including medicine, bioinformatics, economics, speech recognition and computer vision. Example instances of this problem include spotting an object of interest in an image (e.g., a tumor), a pattern of interest in a time-varying signal (e.g., audio analysis), or an object of interest moving in a specific way (e.g., a human's body gesture). Traditional spotting methods, which are based on Dynamic Time Warping or hidden Markov models, use some variant of dynamic programming to register the pattern and the input while accounting for temporal variation between them. At the same time, those methods often suffer from several shortcomings: they may give meaningless solutions when input observations are unreliable or ambiguous, they require a high complexity search across the whole input signal, and they may give incorrect solutions if some patterns appear as smaller parts within other patterns. In this thesis, we develop a framework that addresses these three problems, and evaluate the framework's performance in spotting and recognizing hand gestures in video. The first contribution is a spatiotemporal matching algorithm that extends the dynamic programming formulation to accommodate multiple candidate hand detections in every video frame. The algorithm finds the best alignment between the gesture model and the input, and simultaneously locates the best candidate hand detection in every frame. This allows for a gesture to be recognized even when the hand location is highly ambiguous. The second contribution is a pruning method that uses model-specific classifiers to reject dynamic programming hypotheses with a poor match between the input and model. Pruning improves the efficiency of the spatiotemporal matching algorithm, and in some cases may improve the recognition accuracy. The pruning classifiers are learned from training data, and cross-validation is used to reduce the chance of overpruning. The third contribution is a subgesture reasoning process that models the fact that some gesture models can falsely match parts of other, longer gestures. By integrating subgesture reasoning the spotting algorithm can avoid the premature detection of a subgesture when the longer gesture is actually being performed. Subgesture relations between pairs of gestures are automatically learned from training data. The performance of the approach is evaluated on two challenging video datasets: hand-signed digits gestured by users wearing short sleeved shirts, in front of a cluttered background, and American Sign Language (ASL) utterances gestured by ASL native signers. The experiments demonstrate that the proposed method is more accurate and efficient than competing approaches. The proposed approach can be generally applied to alignment or search problems with multiple input observations, that use dynamic programming to find a solution.

Estimating Sensor Motion from Wide-Field Optical Flow on a Log-Dipolar Sensor

Log-polar image architectures, motivated by the structure of the human visual field, have long been investigated in computer vision for use in estimating motion parameters from an optical flow vector field. Practical problems with this approach have been: (i) dependence on assumed alignment of the visual and motion axes; (ii) sensitivity to occlusion form moving and stationary objects in the central visual field, where much of the numerical sensitivity is concentrated; and (iii) inaccuracy of the log-polar architecture (which is an approximation to the central 20°) for wide-field biological vision. In the present paper, we show that an algorithm based on generalization of the log-polar architecture; termed the log-dipolar sensor, provides a large improvement in performance relative to the usual log-polar sampling. Specifically, our algorithm: (i) is tolerant of large misalignmnet of the optical and motion axes; (ii) is insensitive to significant occlusion by objects of unknown motion; and (iii) represents a more correct analogy to the wide-field structure of human vision. Using the Helmholtz-Hodge decomposition to estimate the optical flow vector field on a log-dipolar sensor, we demonstrate these advantages, using synthetic optical flow maps as well as natural image sequences.

Texture Segregation By Visual Cortex: Perceptual Grouping, Attention, and Learning

A neural model is proposed of how laminar interactions in the visual cortex may learn and recognize object texture and form boundaries. The model brings together five interacting processes: region-based texture classification, contour-based boundary grouping, surface filling-in, spatial attention, and object attention. The model shows how form boundaries can determine regions in which surface filling-in occurs; how surface filling-in interacts with spatial attention to generate a form-fitting distribution of spatial attention, or attentional shroud; how the strongest shroud can inhibit weaker shrouds; and how the winning shroud regulates learning of texture categories, and thus the allocation of object attention. The model can discriminate abutted textures with blurred boundaries and is sensitive to texture boundary attributes like discontinuities in orientation and texture flow curvature as well as to relative orientations of texture elements. The model quantitatively fits a large set of human psychophysical data on orientation-based textures. Object boundar output of the model is compared to computer vision algorithms using a set of human segmented photographic images. The model classifies textures and suppresses noise using a multiple scale oriented filterbank and a distributed Adaptive Resonance Theory (dART) classifier. The matched signal between the bottom-up texture inputs and top-down learned texture categories is utilized by oriented competitive and cooperative grouping processes to generate texture boundaries that control surface filling-in and spatial attention. Topdown modulatory attentional feedback from boundary and surface representations to early filtering stages results in enhanced texture boundaries and more efficient learning of texture within attended surface regions. Surface-based attention also provides a self-supervising training signal for learning new textures. Importance of the surface-based attentional feedback in texture learning and classification is tested using a set of textured images from the Brodatz micro-texture album. Benchmark studies vary from 95.1% to 98.6% with attention, and from 90.6% to 93.2% without attention.