Jane Mulligan

Performance evaluation of color correction approaches for automatic multi-view image and video stitching

Many different automatic color correction approaches have been proposed by different research communities in the past decade. However, these approaches are seldom compared, so their relative performance and applicability are unclear. For multi-view image and video stitching applications, an ideal color correction approach should be effective at transferring the color palette of the source image to the target image, and meanwhile be able to extend the transferred color from the overlapped area to the full target image without creating visual artifacts. In this paper we evaluate the performance of color correction approaches for automatic multi-view image and video stitching. We consider nine color correction algorithms from the literature applied to 40 synthetic image pairs and 30 real mosaic image pairs selected from different applications. Experimental results show that both parametric and non-parametric approaches have members that are effective at transferring colors, while parametric approaches are generally better than non-parametric approaches in extendability.

Trinocular stereo: a real-time algorithm and its evaluation

In telepresence applications each user is immersed in a rendered 3D-world composed from representations transmitted from remote sites. The challenge is to compute dense range data at high frame rates, since participants cannot easily communicate if the processing cycle or network latencies are long. Moreover, errors in new stereoscopic views of the remote 3D-world should be hardly perceptible. To achieve the required speed and accuracy, we use trinocular stereo, a matching algorithm based on the sum of modified normalized cross-correlations, and subpixel disparity interpolation. To increase speed we use Intel IPL functions in the pre-processing steps of background subtraction and image rectification as well as a four-processor parallelization. To evaluate our system we have developed a test-bed which provides a set of registered dense “ground-truth” laser data and image data from multiple views.

Running on empty? The compensatory reserve index.

BACKGROUND Hemorrhage is a leading cause of traumatic death. We hypothesized that state-of-the-art feature extraction and machine learning techniques could be used to discover, detect, and continuously trend beat-to-beat changes in arterial pulse waveforms associated with the progression to hemodynamic decompensation. METHODS We exposed 184 healthy humans to progressive central hypovolemia using lower-body negative pressure to the point of hemodynamic decompensation (systolic blood pressure > 80 mm Hg with or without bradycardia). Initial models were developed using continuous noninvasive blood pressure waveform data. The resulting algorithm calculates a compensatory reserve index (CRI), where 1 represents supine normovolemia and 0 represents the circulatory volume at which hemodynamic decompensation occurs (i.e., “running on empty”). Values between 1 and 0 indicate the proportion of reserve remaining before hemodynamic decompensation—much like the fuel gauge of a car indicates the amount of fuel remaining in the tank. A CRI estimate is produced after the first 30 heart beats, followed by a new CRI estimate after each subsequent beat. RESULTS The CRI model with a 30-beat window has an absolute difference between actual and expected time to decompensation of 0.1, with a SD of 0.09. The model distinguishes individuals with low tolerance to reduced central blood volume (i.e., those most likely to develop early shock) from those with high tolerance and are able to estimate how near or far an individual may be from hemodynamic decompensation. CONCLUSION Machine modeling can quickly and accurately detect and trend central blood volume reduction in real time during the compensatory phase of hemorrhage as well as estimate when an individual is “running on empty” and will decompensate (CRI, 0), well in advance of meaningful changes in traditional vital signs.

View-independent scene acquisition for tele-presence

Tele-immersion is a new medium that enables a user to share a virtual space with remote participants. The user is immersed in a rendered 3D-world that is transmitted from a remote site. To acquire this 3D description we apply bi- and trinocular stereo techniques. The challenge is to compute dense stereo range data at high frame rates, since participants cannot easily communicate if the processing cycle or network latencies are long. Moreover, new views of the received 3D-world must be as accurate as possible. We address both issues of speed and accuracy and we propose a method for combining motion and stereo in order to increase speed and robustness.

Real time trinocular stereo for tele-immersion

Tele-immersion is a technology that augments your space with real-time 3D projections of remote spaces thus facilitating the interaction of people from different places in virtually the same environment. Tele-immersion combines 3D scene recovery from computer vision, and rendering and interaction from computer graphics. We describe the real-time 3D scene acquisition using a new algorithm for trinocular stereo. We extend this method in time by combining motion and stereo in order to increase speed and robustness.

Trinocular stereo for non-parallel configurations

The constraint of a third camera in stereo vision is a useful tool for reducing ambiguity in matching. Most of the systems using trinocular stereo to date however, have used configurations where the image planes of all three cameras are coplanar, or can be rectified to be so. We explore the computation of dense trinocular disparity maps for non-planar camera configurations which arise when cameras surround the object to be modeled. Our approach rectifies the cameras as two independent stereo pairs. We start with an exhaustive lookup scheme and then consider retaining only a list of N disparities per pixel with maximal correlation values for the right and left pairs. Experimental results and comparisons demonstrate that both methods reduce outliers over binocular stereo, and that the N-hypothesis system trades large lookup tables for somewhat lower density of valid matches.

Individual-Specific, Beat-to-beat Trending of Significant Human Blood Loss: The Compensatory Reserve.

ABSTRACT Current monitoring technologies are unable to detect early, compensatory changes that are associated with significant blood loss. We previously introduced a novel algorithm to calculate the Compensatory Reserve Index (CRI) based on the analysis of arterial waveform features obtained from photoplethysmogram recordings. In the present study, we hypothesized that the CRI would provide greater sensitivity and specificity to detect blood loss compared with traditional vital signs and other hemodynamic measures. Continuous noninvasive vital sign waveform data, including CRI, photoplethysmogram, heart rate, blood pressures, SpO2, cardiac output, and stroke volume, were analyzed from 20 subjects before, during, and after an average controlled voluntary hemorrhage of ∼1.2 L of blood. Compensatory Reserve Index decreased by 33% in a linear fashion across progressive blood volume loss, with no clinically significant alterations in vital signs. The receiver operating characteristic area under the curve for the CRI was 0.90, with a sensitivity of 0.80 and specificity of 0.76. In comparison, blood pressures, heart rate, SpO2, cardiac output, and stroke volume had significantly lower receiver operating characteristic area under the curve values and specificities for detecting the same volume of blood loss. Consistent with our hypothesis, CRI detected blood loss and restoration with significantly greater specificity than did other traditional physiologic measures. Single measurement of CRI may enable more accurate triage, whereas CRI monitoring may allow for earlier detection of casualty deterioration.

Outdoor Path Labeling Using Polynomial Mahalanobis Distance

Autonomous robot navigation in outdoor environments remains a challenging and unsolved problem. A key issue is our ability to identify safe or navigable paths far enough ahead of the robot to allow smooth trajectories at acceptable speeds. Colour or texture-based labeling of safe path regions in image sequences is one way to achieve this far field prediction. A challenge for classifiers identifying path and nonpath regions is to make meaningful comparisons of feature vectors at pixels or over a window. Most simple distance metrics cannot use all the information available and therefore the resulting labeling does not tightly capture the visible path. We introduce a new Polynomial Mahalanobis Distance and demonstrate its ability to capture the properties of an initial positive path sample and produce accurate path segmentation with few outliers. Experiments show the method’s effectiveness for path segmentation in natural scenes using both colour and texture feature vectors. The new metric is compared with classifications based on Euclidean and standard Mahalanobis distance and produces superior results.

Detection of Low-volume Blood Loss: Compensatory Reserve Versus Traditional Vital Signs

BACKGROUND Humans are able to compensate for low-volume blood loss with minimal change in traditional vital signs. We hypothesized that a novel algorithm, which analyzes photoplethysmogram (PPG) wave forms to continuously estimate compensatory reserve would provide greater sensitivity and specificity to detect low-volume blood loss compared with traditional vital signs. The compensatory reserve index (CRI) is a measure of the reserve remaining to compensate for reduced central blood volume, where a CRI of 1 represents supine normovolemia and 0 represents the circulating blood volume at which hemodynamic decompensation occurs; values between 1 and 0 indicate the proportion of reserve remaining. METHODS Subjects underwent voluntary donation of 1 U (approximately 450 mL) of blood. Demographic and continuous noninvasive vital sign wave form data were collected, including PPG, heart rate, systolic blood pressure, cardiac output, and stroke volume. PPG wave forms were later processed by the algorithm to estimate CRI values. RESULTS Data were collected from 244 healthy subjects (79 males and 165 females), with a mean (SD) age of 40.1 (14.2) years and mean (SD) body mass index of 25.6 (4.7). After blood donation, CRI significantly decreased in 92% (&agr; = 0.05; 95% confidence interval [CI], 88–95%) of the subjects. With the use of a threshold decrease in CRI of 0.05 or greater for the detection of 1 U of blood loss, the receiver operating characteristic area under the curve was 0.90, with a sensitivity of 0.84 and specificity of 0.86. In comparison, systolic blood pressure (52%; 95% CI, 45–59%), heart rate (65%; 95% CI, 58–72%), cardiac output (47%; 95% CI, 40–54%), and stroke volume (74%; 95% CI, 67–80%) changed in fewer subjects, had significantly lower receiver operating characteristic area under the curve values, and significantly lower specificities for detecting the same volume of blood loss. CONCLUSION Consistent with our hypothesis, CRI detected low-volume blood loss with significantly greater specificity than other traditional physiologic measures. These findings warrant further evaluation of the CRI algorithm in actual trauma settings. LEVEL OF EVIDENCE Diagnostic study, level II.

Local path planning in image space for autonomous robot navigation in unstructured environments

An approach to stereo based local path planning in unstructured environments is presented. The approach differs from previous stereo based and image based planning systems (e.g. top-down occupancy grid planners, autonomous highway driving algorithms, and view-sequenced route representation), in that it uses specialized cost functions to find paths through an occupancy grid representation of the world directly in the image plane and forgoes a projection of cost information from the image plane down onto a top-down 2D Cartesian cost map. We discuss three cost metrics for path selection in image space. We present a basic image based planning system, discuss its susceptibility to rotational and translational oscillation, and present and implement two extensions to the basic system that overcome these limitations - a cylindrical based image system and a hierarchical planning system. All three systems are implemented in an autonomous robot and are tested against a standard top-down 2D Cartesian planning system on three outdoor courses of varying difficulty. We find that the basic image based planning system fails under certain conditions; however, the cylindrical based system is well suited to the task of local path planning and for use as a high resolution local planning component of a hierarchical planning system.