Dario Comanducci

Robust tracking and remapping of eye appearance with passive computer vision

A single-camera iris-tracking and remapping approach based on passive computer vision is presented. Tracking is aimed at obtaining accurate and robust measurements of the iris/pupil position. To this purpose, a robust method for ellipse fitting is used, employing search constraints so as to achieve better performance with respect to the standard RANSAC algorithm. Tracking also embeds an iris localization algorithm (working as a bootstrap multiple-hypotheses generation step), and a blink detector that can detect voluntary eye blinks in human-computer interaction applications. On-screen remapping incorporates a head-tracking method capable of compensating for small user-head movements. The approach operates in real time under different light conditions and in the presence of distractors. An extensive set of experiments is presented and discussed. In particular, an evaluation method for the choice of layout of both hardware components and calibration points is described. Experiments also investigate the importance of providing a visual feedback to the user, and the benefits gained from performing head compensation, especially during image-to-screen map calibration.

Accurate Automatic Localization of Surfaces of Revolution for Self-Calibration and Metric Reconstruction

In this paper, we address the problem of the automatic metric reconstruction Surface of Revolution (SOR) from a single uncalibrated view. The apparent contour and the visible portions of the imaged SOR cross sections are extracted and classified. The harmonic homology that models the image projection of the SOR is also estimated. The special care devoted to accuracy and robustness with respect to outliers makes the approach suitable for automatic camera calibration and metric reconstruction from single uncalibrated views of a SOR. Robustness and accuracy are obtained by embedding a graph-based grouping strategy (Euclidean Minimum Spanning Tree) into an Iterative Closest Point framework for projective curve alignment at multiple scales. Classification of SOR curves is achieved through a 2-dof voting scheme based on a pencil of conics novel parametrization. The main contribution of this work is to extend the domain of automatic single view reconstruction from piecewise planar scenes to scenes including curved surfaces, thus allowing to create automatically realistic image models of man-made objects. Experimental results with real images taken from the internet are reported, and the effectiveness and limitations of the approach are discussed.

Automatic bus line number localization and recognition on mobile phones: a computer vision aid for the visually impaired

In this paper, machine learning and geometric computer vision are combined for the purpose of automatic reading bus line numbers with a smart phone. This can prove very useful to improve the autonomy of visually impaired people in urban scenarios. The problem is a challenging one, since standard geometric image matching methods fail due to the abundance of distractors, occlusions, illumination changes, highlights and specularities, shadows, and perspective distortions. The problem is solved by locating the main geometric entities of the bus facade through a cascade of classifiers, and then refining the matching with robust geometric matching. The method works in real time and, as experimental results show, has a good performance in terms of recognition rate and reliability.

Camera calibration with two arbitrary coaxial circles

We present an approach for camera calibration from the image of at least two circles arranged in a coaxial way. Such a geometric configuration arises in static scenes of objects with rotational symmetry or in scenes including generic objects undergoing rotational motion around a fixed axis. The approach is based on the automatic localization of a surface of revolution (SOR) in the image, and its use as a calibration artifact. The SOR can either be a real object in a static scene, or a “virtual surface” obtained by frame superposition in a rotational sequence. This provides a unified framework for calibration from single images of SORs or from turntable sequences. Both the internal and external calibration parameters (square pixels model) are obtained from two or more imaged cross sections of the SOR, whose apparent contour is also exploited to obtain a better calibration accuracy. Experimental results show that this calibration approach is accurate enough for several vision applications, encompassing 3D realistic model acquisition from single images, and desktop 3D object scanning.

Shape reconstruction and texture sampling by active rectification and virtual view synthesis

In this paper it is shown how to obtain a three-dimensional, textured object model by relying exclusively on image warping 2D-2D transformations. To achieve this goal, a dual (laser and natural light) illumination of the scene is exploited. Shape reconstruction is not based on triangulation, but on the planar rectification and collation of laser profiles. Texture sampling uses laser profile warping and compositing rather than 3D-2D model shape reprojection, so as to synthesize a virtual view of the object at hand, that is matched against the natural light sequence to obtain color data. Shape reconstruction and texture sampling are independent from each other, and can be run in parallel, thus speeding up the process. Moreover, the approach requires only a simple manual setup, and is effective with textured or textureless objects of any shape. Experimental results demonstrate that the approach combines the high accuracy of state-of-the-art active reconstruction methods with the flexibility of uncalibrated methods, improving on both. Specifically, avoiding any intermediate 3D measurement - in particular the external camera calibration parameters - has a dramatic impact on both model shape an texture accuracy, and also adds robustness w.r.t. any estimation errors of internal camera parameters.

A Desktop 3D Scanner Exploiting Rotation and Visual Rectification of Laser Profiles

We describe a low cost system for metric 3D scanning from uncalibrated images based on rotational kinematic constraints. The system is composed by a turntable, an offthe- shelf camera and a laser stripe illuminator. System operation is based on the construction of the virtual image of a surface of revolution (SOR), from which two imaged SOR cross-sections are obtained in an automatic way, and internal camera calibration is performed by exploiting the same object being scanned. Shape acquisition is finally obtained by laser profile rectification and collation. Experiments with real data are shown, providing an insight into both camera calibration and shape reconstruction performance. System accuracy appears to be adequate for desktop applications.

Robust Iris Localization and Tracking based on Constrained Visual Fitting

A robust iris localization and tracking algorithm based on computer vision is presented. The iris localization algorithm acts as a bootstrap for the tracking algorithm, providing it with a set of multiple hypotheses to restart from in the case of a tracking failure. Tracking is performed with a RANSAC-like robust method for ellipse fitting that incorporates search constraints so as to increase the overall accuracy with respect to the standard RANSAC approach. Experimental results show that the algorithm is fast, accurate and robust enough for applications in the field of human-machine interaction, being particularly suitable for users with severe motor disabilities.

Low-Cost 3D Scanning by Exploiting Virtual Image Symmetries

We present a low-cost, hybrid active/passive 3D scanning system based on an off-the-shelf camera, a laser stripe illuminator and a turntable. The system combines the good accuracy of active triangulation approaches with the flexibility of self-calibration based approaches. System operation is based on the construction of a single view of a virtual surface of revolution, from which camera calibration (both intrinsic and extrinsic parameters) is performed by exploiting the same object being scanned. Shape acquisition is finally obtained by laser profile reconstruction and collation. Experiments with both synthetic and real data are shown, providing an insight into both camera calibration and shape reconstruction performance.

Vision-Based magnification of corneal endothelium frames

We present a fast and effective method to compute a high-resolution image of the corneal endothelium starting from a low-resolution video sequence obtained with a general purpose biomicroscope. Our goal is to exploit information redundancy in the sequence so as to achieve via software a magnification power and an image quality typical of dedicated hardware, such as the confocal microscope. The method couples SVM training with graph-based registration, and explicitly takes into account the characteristics of the application domain. Results on long, real sequences and comparative tests against general-purpose super-resolution approaches are presented and discussed.

Behavior monitoring through automatic analysis of video sequences

This paper addresses the problem of classifying actions performed by a human subject in a video sequence. A representation eigenspace approach based on the visual appearance is used to train the classifier. Before dimensionality reduction exploiting the PCA/LLE algorithms, a high dimensional description of each frame of the video sequence is constructed, based on foreground blob analysis. The classification task is performed by matching incrementally the reduced representation of the test image sequence against each of the learned ones, and accumulating matching scores until a decision is obtained; to this aim, two different metrics are introduced and evaluated. Experimental results demonstrate that the approach is accurate enough and feasible for behavior classification. Furthermore, we argue that the choice of both the feature descriptor and the metric for the matching score can dramatically influence the performance of the results.