David Corrigan

Bio: David Corrigan is an academic researcher from Trinity College, Dublin. The author has contributed to research in topics: Image restoration & Motion estimation. The author has an hindex of 9, co-authored 29 publications receiving 310 citations. Previous affiliations of David Corrigan include Huawei.

Video Matting Using Motion Extended GrabCut

Abstract: GrabCut is perhaps the most powerful semi-automatic algorithm for matting presented to date. In its existing form, it is not suitable for video object segmentation. This paper considers major extensions that make it suitable for this purpose. A method for initialising matting without user intervention is presented, followed by a more robust data model using a Mean Shift algorithm to control model complexity. In addition, normalised motion information as well as colour is used to form joint colour and motion feature vectors. This improves the robustness of the mattes in the presence of colour camouflage and decreases the user intervention required for a successful result. Comparison between GrabCut and the proposed Motion Extended GrabCut (MxGrabCut), shows the improvement for video matting.

3D Object Recognition Based on Volumetric Representation Using Convolutional Neural Networks

Abstract: Following the success of Convolutional Neural Networks on object recognition and image classification using 2D images; in this work the framework has been extended to process 3D data. However, many current systems require huge amount of computation cost for dealing with large amount of data. In this work, we introduce an efficient 3D volumetric representation for training and testing CNNs and we also build several datasets based on the volumetric representation of 3D digits, different rotations along the x, y and z axis are also taken into account. Unlike the normal volumetric representation, our datasets are much less memory usage. Finally, we introduce a model based on the combination of CNN models, the structure of the model is based on the classical LeNet. The accuracy result achieved is beyond the state of art and it can classify a 3D digit in around 9 ms.

A Video Database for the Development of Stereo-3D Post-Production Algorithms

Abstract: This paper introduces a new database of freely available stereo-3D content designed to facilitate research in stereo post-production. It describes the structure and content of the database and provides some details about how the material was gathered. The database includes examples of many of the scenarios characteristic to broadcast footage. Material was gathered at different locations including a studio with controlled lighting and both indoor and outdoor on-location sites with more restricted lighting control. An intended consequence of gathering the material is that the database contains examples of degradations that would be commonly present in real-world scenarios. This paper describes one such artefact caused by uneven exposure in the stereo views, causing saturation in the over-exposed view. An algorithm is proposed that replaces the saturated data by interpolating data from the unsaturated view in the wavelet domain.

Convolutional Neural Network for 3D object recognition using volumetric representation

Abstract: Following the success of Convolutional Neural Networks (CNNs) on object recognition using 2D images, they are extended in this paper to process 3D data. Nearly most of current systems require huge amount of computation for dealing with large amount of data. In this paper, an efficient 3D volumetric object representation, Volumetric Accelerator (VOLA), is presented which requires much less memory than the normal volumetric representations. On this basis, a few 3D digit datasets using 2D MNIST and 2D digit fonts with different rotations along the x, y, and z axis are introduced. Finally, we introduce a combination of multiple CNN models based on the famous LeNet model. The trained CNN models based on the generated dataset have achieved the average accuracy of 90.30% and 81.85% for 3D-MNIST and 3D-Fonts datasets, respectively. Experimental results show that VOLA-based CNNs perform 1.5x faster than the original LeNet.

The perception of egocentric distances in virtual environments - A review

Abstract: Over the last 20 years research has been done on the question of how egocentric distances, i.e., the subjectively reported distance from a human observer to an object, are perceived in virtual environments. This review surveys the existing literature on empirical user studies on this topic. In summary, there is a mean estimation of egocentric distances in virtual environments of about 74p of the modeled distances. Many factors possibly influencing distance estimates were reported in the literature. We arranged these factors into four groups, namely measurement methods, technical factors, compositional factors, and human factors. The research on these factors is summarized, conclusions are drawn, and promising areas for future research are outlined.

Distance perception in NPR immersive virtual environments, revisited

Abstract: Non-photorealistic rendering (NPR) is a representational technique that allows communicating the essence of a design while giving the viewer the sense that the design is open to change. Our research aims to address the question of how to effectively use non-photorealistic rendering in immersive virtual environments to enable the intuitive exploration of early architectural design concepts at full scale. Previous studies have shown that people typically underestimate egocentric distances in immersive virtual environments, regardless of rendering style, although we have recently found that distance estimation errors are minimized in the special case that the virtual environment is a high-fidelity replica of a real environment that the viewer is presently in or has recently been in. In this paper we re-examine the impact of rendering style on distance perception accuracy in this virtual environments context. Specifically, we report the results of an experiment that seeks to assess the accuracy with which people judge distances in a non-photorealistically rendered virtual environment that is a directly-derived stylistic abstraction of the actual environment that they are currently in. Our results indicate that people tend to underestimate distances to a significantly greater extent in a co-located virtual environment when it is rendered using a line-drawing style than when it is rendered using high fidelity textures derived from photographs.

The effects of virtual reality, augmented reality, and motion parallax on egocentric depth perception

Abstract: As the use of virtual and augmented reality applications becomes more common, the need to fully understand how observers perceive spatial relationships grows more critical. One of the key requirements in engineering a practical virtual or augmented reality system is accurately conveying depth and layout. This requirement has frequently been assessed by measuring judgments of egocentric depth. These assessments have shown that observers in virtual reality (VR) perceive virtual space as compressed relative to the real-world, resulting in systematic underestimations of egocentric depth. Previous work has indicated that similar effects may be present in augmented reality (AR) as well.This paper reports an experiment that directly measured egocentric depth perception in both VR and AR conditions; it is believed to be the first experiment to directly compare these conditions in the same experimental framework. In addition to VR and AR, two control conditions were studied: viewing real-world objects, and viewing real-world objects through a head-mounted display. Finally, the presence and absence of motion parallax was crossed with all conditions. Like many previous studies, this one found that depth perception was underestimated in VR, although the magnitude of the effect was surprisingly low. The most interesting finding was that no underestimation was observed in AR.

Egocentric depth judgments in optical, see-through augmented reality

Abstract: A fundamental problem in optical, see-through augmented reality (AR) is characterizing how it affects the perception of spatial layout and depth. This problem is important because AR system developers need to both place graphics in arbitrary spatial relationships with real-world objects, and to know that users will perceive them in the same relationships. Furthermore, AR makes possible enhanced perceptual techniques that have no real-world equivalent, such as x-ray vision, where AR users are supposed to perceive graphics as being located behind opaque surfaces. This paper reviews and discusses protocols for measuring egocentric depth judgments in both virtual and augmented environments, and discusses the well-known problem of depth underestimation in virtual environments. It then describes two experiments that measured egocentric depth judgments in AR. Experiment I used a perceptual matching protocol to measure AR depth judgments at medium and far-field distances of 5 to 45 meters. The experiment studied the effects of upper versus lower visual field location, the x-ray vision condition, and practice on the task. The experimental findings include evidence for a switch in bias, from underestimating to overestimating the distance of AR-presented graphics, at ~ 23 meters, as well as a quantification of how much more difficult the x-ray vision condition makes the task. Experiment II used blind walking and verbal report protocols to measure AR depth judgments at distances of 3 to 7 meters. The experiment examined real-world objects, real-world objects seen through the AR display, virtual objects, and combined real and virtual objects. The results give evidence that the egocentric depth of AR objects is underestimated at these distances, but to a lesser degree than has previously been found for most virtual reality environments. The results are consistent with previous studies that have implicated a restricted field-of-view, combined with an inability for observers to scan the ground plane in a near-to-far direction, as explanations for the observed depth underestimation.

The effects of head-mounted display mechanical properties and field of view on distance judgments in virtual environments

Abstract: Research has shown that people are able to judge distances accurately in full-cue, real-world environments using visually directed actions. However, in virtual environments viewed with head-mounted display (HMD) systems, there is evidence that people act as though the virtual space is smaller than intended. This is a surprising result given how well people act in real environments. The behavior in the virtual setting may be linked to distortions in the available visual cues or to a person's ability to locomote without vision. Either could result from issues related to added mass, moments of inertia, and restricted field of view in HMDs. This article describes an experiment in which distance judgments based on normal real-world and HMD viewing are compared with judgments based on real-world viewing while wearing two specialized devices. One is a mock HMD, which replicated the mass, moments of inertia, and field of view of the HMD and the other an inertial headband designed to replicate the mass and moments of inertia of the HMD, but constructed to not restrict the field of view of the observer or otherwise feel like wearing a helmet. Distance judgments using the mock HMD showed a statistically significant underestimation relative to the no restriction condition but not of a magnitude sufficient to account for all the distance compression seen in the HMD. Indicated distances with the inertial headband were not significantly smaller than those made with no restrictions.