People use orientation of their faces to convey rich, inter-personal information. For example, a person will direct his face to indicate who the intended target of the conversation is. Similarly in a conversation, face orientation is a non-verbal cue to listener when to switch role and start speaking, and a nod indicates that a person has understands, or agrees with, what is being said. Further more, face pose estimation plays an important role in human-computer interaction, virtual reality applications, human behaviour analysis, pose-independent face recognition, driver s vigilance assessment, gaze estimation, etc. Robust face recognition has been a focus of research in computer vision community for more than two decades. Although substantial research has been done and numerous methods have been proposed for face recognition, there remain challenges in this field. One of these is face recognition under varying poses and that is why face pose estimation is still an important research area. In computer vision, face pose estimation is the process of inferring the face orientation from digital imagery. It requires a serious of image processing steps to transform a pixel-based representation of a human face into a high-level concept of direction. An ideal face pose estimator should be invariant to a variety of image-changing factors such as camera distortion, lighting condition, skin colour, projective geometry, facial hairs, facial expressions, presence of accessories like glasses and hats, etc. Face pose estimation has been a focus of research for about two decades and numerous research contributions have been presented in this field. Face pose estimation techniques in literature have still some shortcomings and limitations in terms of accuracy, applicability to monocular images, being autonomous, identity and lighting variations, image resolution variations, range of face motion, computational expense, presence of facial hairs, presence of accessories like glasses and hats, etc. These shortcomings of existing face pose estimation techniques motivated the research work presented in this thesis. The main focus of this research is to design and develop novel face pose estimation algorithms that improve automatic face pose estimation in terms of processing time, computational expense, and invariance to different conditions.

Visual Simultaneous Localization And Mapping (SLAM) allows for three dimensionalreconstruction from a camera’s output and simultaneous positioning of the camera withinthe reconstruction. With use cases ranging from autonomous vehicles to augmentedreality, the SLAM field has garnered interest both commercially and academically. A SLAM system performs odometry as it estimates the camera’s movement throughthe scene. The incremental estimation of odometry is not error free and exhibits driftover time with map inconsistencies as a result. Detecting the return to a previously seenplace, a loop, means that this new information regarding our position can be incorporatedto correct the trajectory retroactively. Loop detection can also facilitate relocalization ifthe system loses tracking due to e.g. heavy motion blur. This thesis proposes an odometric system making use of bundle adjustment within akeyframe based stereo SLAM application. This system is capable of detecting loops byutilizing the algorithm FAB-MAP. Two aspects of this system is evaluated, the odometryand the capability to relocate. Both of these are evaluated using the EuRoC MAV dataset,with an absolute trajectory RMS error ranging from 0.80 m to 1.70 m for the machinehall sequences. The capability to relocate is evaluated using a novel methodology that intuitively canbe interpreted. Results are given for different levels of strictness to encompass differentuse cases. The method makes use of reprojection of points seen in keyframes to definewhether a relocalization is possible or not. The system shows a capability to relocate inup to 85% of all cases when a keyframe exists that can project 90% of its points intothe current view. Errors in estimated poses were found to be correlated with the relativedistance, with errors less than 10 cm in 23% to 73% of all cases. The evaluation of the whole system is augmented with an evaluation of local imagedescriptors and pose estimation algorithms. The descriptor SIFT was found to performbest overall, but demanding to compute. BRISK was deemed the best alternative for afast yet accurate descriptor. Conclusions that can be drawn from this thesis is that FAB-MAP works well fordetecting loops as long as the addition of keyframes is handled appropriately.

This dissertation addresses a number of inter-related and fundamental problems in computer vision. Specifically, we address object discovery, recognition, segmentation, and 3D pose estimation in images, as well as 3D scene reconstruction and scene interpretation. The key ideas behind our approaches include using shape as a basic object feature, and using structured prediction modeling paradigms for representing objects and scenes. In this work, we make a number of new contributions both in computer vision and machine learning. We address the vision problems of shape matching, shape-based mining of objects in arbitrary image collections, context-aware object recognition, monocular estimation of 3D object poses, and monocular 3D scene reconstruction using shape from texture. Our work on shape-based object discovery is the first to show that meaningful objects can be extracted from a collection of arbitrary images, without any human supervision, by shape matching. We also show that a spatial repetition of objects in images (e.g., windows on a building facade, or cars lined up along a street) can be used for 3D scene reconstruction from a single image. The aforementioned topics have never been addressed in the literature. The dissertation also presents new algorithms and object representations for the aforementioned vision problems. We fuse two traditionally different modeling paradigms Conditional Random Fields (CRF) and Random Forests (RF) into a unified framework, referred to as (RF)^2. We also derive theoretical error bounds of estimating distribution ratios by a two-class RF, which is then used to derive the theoretical performance bounds of a two-class (RF)^2. Thorough experimental evaluation of individual aspects of all our approaches is presented. In general, the experiments demonstrate that we outperform the state of the art on the benchmark datasets, without increasing complexity and supervision in training. / Graduation date: 2011 / Access restricted to the OSU Community at author's request from May 12, 2011 - May 12, 2012

We present an image-based approach to infer 3D structure parameters using a probabilistic "shape+structure'' model. The 3D shape of a class of objects may be represented by sets of contours from silhouette views simultaneously observed from multiple calibrated cameras. Bayesian reconstructions of new shapes can then be estimated using a prior density constructed with a mixture model and probabilistic principal components analysis. We augment the shape model to incorporate structural features of interest; novel examples with missing structure parameters may then be reconstructed to obtain estimates of these parameters. Model matching and parameter inference are done entirely in the image domain and require no explicit 3D construction. Our shape model enables accurate estimation of structure despite segmentation errors or missing views in the input silhouettes, and works even with only a single input view. Using a dataset of thousands of pedestrian images generated from a synthetic model, we can perform accurate inference of the 3D locations of 19 joints on the body based on observed silhouette contours from real images.

Alignment is a prevalent approach for recognizing 3D objects in 2D images. A major problem with current implementations is how to robustly handle errors that propagate from uncertainties in the locations of image features. This thesis gives a technique for bounding these errors. The technique makes use of a new solution to the problem of recovering 3D pose from three matching point pairs under weak-perspective projection. Furthermore, the error bounds are used to demonstrate that using line segments for features instead of points significantly reduces the false positive rate, to the extent that alignment can remain reliable even in cluttered scenes.

In this text, we present two stereo-based head tracking techniques along with a fast 3D model acquisition system. The first tracking technique is a robust implementation of stereo-based head tracking designed for interactive environments with uncontrolled lighting. We integrate fast face detection and drift reduction algorithms with a gradient-based stereo rigid motion tracking technique. Our system can automatically segment and track a user's head under large rotation and illumination variations. Precision and usability of this approach are compared with previous tracking methods for cursor control and target selection in both desktop and interactive room environments. The second tracking technique is designed to improve the robustness of head pose tracking for fast movements. Our iterative hybrid tracker combines constraints from the ICP (Iterative Closest Point) algorithm and normal flow constraint. This new technique is more precise for small movements and noisy depth than ICP alone, and more robust for large movements than the normal flow constraint alone. We present experiments which test the accuracy of our approach on sequences of real and synthetic stereo images. The 3D model acquisition system we present quickly aligns intensity and depth images, and reconstructs a textured 3D mesh. 3D views are registered with shape alignment based on our iterative hybrid tracker. We reconstruct the 3D model using a new Cubic Ray Projection merging algorithm which takes advantage of a novel data structure: the linked voxel space. We present experiments to test the accuracy of our approach on 3D face modelling using real-time stereo images.

Face detection and pose estimation are two widely studied problems - mainly because of their use as subcomponents in important applications, e.g. face recognition. In this thesis I investigate a new approach to the general problem of object detection and pose estimation and apply it to faces. Face detection can be considered a special case of this general problem, but is complicated by the fact that faces are non-rigid objects. The basis of the new approach is the use of scale and orientation invariant feature structures - feature triplets - extracted from the image, as well as a biologically inspired associative structure which maps from feature triplets to desired responses (position, pose, etc.). The feature triplets are constructed from curvature features in the image and coded in a way to represent distances between major facial features (eyes, nose and mouth). The final system has been evaluated on different sets of face images.

Computation of the position and orientation of an object with respect to a camera from its images is called pose estimation problem. Pose estimation is one of the major problems in computer vision, robotics and photogrammetry. Object tracking, object recognition, self-localization of robots are typical examples for the use of pose estimation. Determining the pose of an object from its projections requires 3D model of an object in its own reference system, the camera parameters and 2D image of the object. Most of the pose estimation algorithms require the correspondences between the 3D model points of the object and 2D image points. In this study, four well-known pose estimation algorithms requiring the 2D-3D correspondences to be known a priori / namely, Orthogonal Iterations, POSIT, DLT and Efficient PnP are compared. Moreover, two other well-known algorithms that solve the correspondence and pose problems simultaneously / Soft POSIT and Blind- PnP are also compared in the scope of this thesis study. In the first step of the simulations, synthetic data is formed using a realistic motion scenario and the algorithms are compared using this data. In the next step, real images captured by a calibrated camera for an object with known 3D model are exploited. The simulation results indicate that POSIT algorithm performs the best among the algorithms requiring point correspondences. Another result obtained from the experiments is that Soft-POSIT algorithm can be considered to perform better than Blind-PnP algorithm.

<p>Face detection and pose estimation are two widely studied problems - mainly because of their use as subcomponents in important applications, e.g. face recognition. In this thesis I investigate a new approach to the general problem of object detection and pose estimation and apply it to faces. Face detection can be considered a special case of this general problem, but is complicated by the fact that faces are non-rigid objects. The basis of the new approach is the use of scale and orientation invariant feature structures - feature triplets - extracted from the image, as well as a biologically inspired associative structure which maps from feature triplets to desired responses (position, pose, etc.). The feature triplets are constructed from curvature features in the image and coded in a way to represent distances between major facial features (eyes, nose and mouth). The final system has been evaluated on different sets of face images.</p>

<p>This thesis is about channel-coded feature maps applied in view-based object recognition, tracking, and machine learning. A channel-coded feature map is a soft histogram of joint spatial pixel positions and image feature values. Typical useful features include local orientation and color. Using these features, each channel measures the co-occurrence of a certain orientation and color at a certain position in an image or image patch. Channel-coded feature maps can be seen as a generalization of the SIFT descriptor with the options of including more features and replacing the linear interpolation between bins by a more general basis function.</p><p>The general idea of channel coding originates from a model of how information might be represented in the human brain. For example, different neurons tend to be sensitive to different orientations of local structures in the visual input. The sensitivity profiles tend to be smooth such that one neuron is maximally activated by a certain orientation, with a gradually decaying activity as the input is rotated.</p><p>This thesis extends previous work on using channel-coding ideas within computer vision and machine learning. By differentiating the channel-coded feature maps with respect to transformations of the underlying image, a method for image registration and tracking is constructed. By using piecewise polynomial basis functions, the channel coding can be computed more efficiently, and a general encoding method for N-dimensional feature spaces is presented.</p><p>Furthermore, I argue for using channel-coded feature maps in view-based pose estimation, where a continuous pose parameter is estimated from a query image given a number of training views with known pose. The optimization of position, rotation and scale of the object in the image plane is then included in the optimization problem, leading to a simultaneous tracking and pose estimation algorithm. Apart from objects and poses, the thesis examines the use of channel coding in connection with Bayesian networks. The goal here is to avoid the hard discretizations usually required when Markov random fields are used on intrinsically continuous signals like depth for stereo vision or color values in image restoration.</p><p>Channel coding has previously been used to design machine learning algorithms that are robust to outliers, ambiguities, and discontinuities in the training data. This is obtained by finding a linear mapping between channel-coded input and output values. This thesis extends this method with an incremental version and identifies and analyzes a key feature of the method -- that it is able to handle a learning situation where the correspondence structure between the input and output space is not completely known. In contrast to a traditional supervised learning setting, the training examples are groups of unordered input-output points, where the correspondence structure within each group is unknown. This behavior is studied theoretically and the effect of outliers and convergence properties are analyzed.</p><p>All presented methods have been evaluated experimentally. The work has been conducted within the cognitive systems research project COSPAL funded by EC FP6, and much of the contents has been put to use in the final COSPAL demonstrator system.</p>