 |
Eye Detection Using Composite Features
Chunghoon Kim
We
propose a new discriminant analysis using composite vectors for eye
detection. A composite vector consists of a number of pixels inside a
window on an image. The proposed C-BDA is a biased discriminant
analysis using the covariance of composite vectors. In the hybrid
cascade detector constructed for eye detection, Haar-like features are
used in the earlier stages and composite features obtained from C-BDA
are used in the later stages. The experimental results show that the
proposed detector provides robust performance to several kinds of
variations.
|
935 |
 |
Random Pattern Backgrounds for 3D Shape Reconstruction
Masahiro Toyoura
Correct
silhouettes are required for 3D shape reconstruction from images. The
silhouettes indicate where an object is in images. We propose a
silhouette extraction method from images obtained from multiple
cameras. Even for objects in unknown color, the method can be adopted.
The method is realized with our proposed random pattern background. The
random pattern has many small regions with randomly-selected colors. By
using the random pattern backgrounds, we can keep the rate of missing
parts below a specified percentage.
|
935 |
 |
Multiflash Imaging and Applications
Daniel Vaquero
We
describe a method for acquiring high-quality depth discontinuity maps
from real-world scenes. Our approach is based on a camera with multiple
flashes, which enables the detection of depth edges without performing
a full 3D reconstruction of the scene. The depth edge map is useful in
several applications in computer vision and graphics, such as
non-photorealistic rendering, medical imaging, stereo vision and
fingerspelling recognition. Recent advances on the placement of light
sources and the detection of depth edges in dynamic scenes are also
presented.
|
935 |
 |
Object Contour Detection Using Structured Light
Juneho Yi
This
research describes a novel approach to object contour detection using
structured light. We strategically project structured light and exploit
distortion of the light pattern in the structured light image along
depth discontinuities to reliably detect depth edges. We present a
method that guarantees detection of depth edges for a given range of
object location with an accurate control of key parameters such as
detectable depth difference and stripe width. Experimental results show
that the proposed method accurately detects depth edges of shapes of
human hands and bodies as well as general objects.
|
935 |
 |
Classification of Humans and their Activity
Justin Muncaster
In
this project we attempt to create a system capable of automatically
detecting, tracking, and classifying behavior of actors in a scene.
Towards this end we consider the system as a whole. We seek to identify
features for identifying humans and algorithms for classifying their
activity.
|
935 |
 |
Implicit 3D Modeling and Tracking for Anywhere Augmentation
Sehwan Kim
We
present an online 3D modeling and tracking methodology that uses aerial
photographs for mobile augmented reality. Instead of relying on models
which are created in advance, the system generates a 3D model for a
real building on the fly by combining frontal and aerial views with the
help of an optical sensor, an inertial sensor, a GPS unit and a few
mouse clicks. An unscented Kalman filter (UKF) approach is implemented
as a camera pose estimator.
|
935 |
 |
An Integrated System for Moving Object Classification in Surveillance Videos
Longbin Chen
Moving
object classification in far-field video is a key component of smart
surveillance systems. In this paper, we propose a reliable system for
person-vehicle classification which works well in challenging
real-world conditions, including the presence of shadows, low
resolution imagery, perspective distortions, arbitrary camera
viewpoints, and groups of people. Our system runs in real-time (30Hz)
on conventional machines and has low memory consumption. We achieved
accurate results by relying on powerful discriminative features,
including a novel measure of object deformation based on differences of
histograms of oriented gradients. We also provide an interactive user
interface, enabling users to specify regions of interest for each class
and correct for perspective distortions by specifying different sizes
in different positions of the camera view. Finally, we use an automatic
adaptation process to continuously update the parameters of the system
so that its performance increases for a particular environment.
Experimental results demonstrate the effectiveness of our system in
standard dataset and a variety of video clips captured with our
surveillance cameras.
|
935 |
 |
Efficient Partial Shape Matching Using Smith-Waterman Algorithm
Longbin Chen
We
present an efficient partial shape matching method based on the
Smith-Waterman algorithm. For two contours of m and n points
respectively, the complexity of our method to find similar parts is
only O(mn). In addition to this improvement in efficiency, we also
obtain comparable accurate matching with fewer shape descriptors. Also,
in contrast to arbitrary distance functions that are used by previous
methods, we use a probabilistic similarity measurement, p-value, to
evaluate the similarity of two shapes. Our experiments on several
public shape databases indicate that our method outperforms
state-of-the-art global and partial shape matching algorithms in
various scenarios.
|
935 |
 |
Subdivision Graphs And Web Visualizations
Brynjar Gretarsson
The
goal of this project is to make it easier for data analysts to
visualize and interact with large graphs of data. At the core of our
layout algorithm is a mass spring model, but we have made many
optimizations to enable visualization of tens of thousands of nodes. We
use octrees to reduce the number of repelling forces and we also use
multiple levels of abstraction to improve both the layout algorithm and
the readability of the graph. Recently we have been looking into ways
of creating web visualizations wich can run natively in any web browser.
|
Elings |
 |
Real-time Graph Manipulation with Hundreds of Thousands of Nodes
Peterson Trethewey
We present two original techniques for real time graph layout
manipulation inspired by methods from the fields of 3D mesh
deformation and static graph drawing. Our software takes a large
graph as input, and then allows a user to explore the graph, extract
information, make the layout more aesthetic, and even generate, from
scratch, meaningful representations interactively in real time. Our
techniques are designed for use on graphs with no multiple edges or
self loops and run at interactive rates on graphs of hundreds of
thousands of nodes.
|
Elings |
 |
3D Panoramas for Outdoor Augmented Reality
Chris Coffin, Jonathan Ventura and Jason Wither
The
focus of this project is creating 3D panoramas semi-automatically in
real time. We attach a laser range finder to a camera to simultaneously
acquire depth samples with images. While panning the camera around the
environment to build the panorama, we use image segmentation to "cut
out" 3D objects. In collaboration with an expert remote user, the
system allows for interactive 3D modeling and accurate annotations.
|
Elings |
 |
Input Devices for Spherical Environments
Svetlin Bostandjiev
This
project aims to find the best two dimensional input device for
spherical displays. Our hypothesis is that a pointing device such as a
laser pointer would outperform the traditional mouse due to the shape
of the display. Our tests are conducted in the UCSB Allosphere, a novel
large-scale environment, which serves as an instrument for immersive
visualization and simulation. The Allosphere will potentially be one of
the world's largest immersive environments.
|
Elings |
 |
InViTe - Distributed Augmented Reality
Chris Coffin
Distributed
augmented reality has the goal of immersing the user in a remote
location using images or video taken at that location, often in absence
of a virtual model. A simple example is Google's street view in which
the user is able to explore a series of panoramas. InViTe stands for
Interactive Video Telecollaboration and focuses on providing an
immersive distributed augmented reality experience while relying
primarily on data from video cameras. We start with a simple passive
television like view through a remote camera and extend the interaction
to the point where a user is able to virtually reach through their own
screen and into a remote location.
|
Elings |
 |
Enhancing Classroom and Distance Learning Through Augmented Reality
Christopher Coffin and Svetlin Bostandjiev
We
present a multimedia solution for easily adding virtual annotations to
class lectures through the use of augmented videoconferencing and
tracked physical props. In the classroom, the actions of the instructor
are captured by one or more cameras. We then use a normal desktop
computer to add virtual data to the camera image. Our software solution
tracks the physical objects and allows for overlays of relevant
information, optionally deriving information from the movement of the
objects. Additionally, our solution allows students to interact with
the instructor and virtual data through the augmented video, even when
distributed over the Internet.
|
Elings |
 |
Evaluating Display Types for AR Selection and Annotation
Jason Wither
We
evaluate different display devices for selection or annotation tasks in
augmented reality (AR). We compare a head-mounted display and two
hand-held displays with two tasks, a search task and navigation task.
We found that using a hand-held display in a "magic lens" configuration
was faster for cursor movement than either of the other two displays.
There was no significant difference among the displays regarding search
time for physical or virtual objects. |
Elings |
 |
Interactive Stereoscopic FogScreen: Towards a 3D Walkthrough Display
Cha Lee
We are working towards a 3D walk-through
display based on the FogScreen, an "immaterial" indoor 2D
projection screen. We extended the basic 2D FogScreen setup with
head tracking to provide correct perspective rendering for a
single user. Second, we evaluated multiple types of
stereoscopic imagery. Third, we presented the front and back views
of a graphics scene on the two sides of the FogScreen, so that
one can cross the screen to see the content from the back.
Fourth, we project imagery on multiple planes (here: the
FogScreen and a wall), such that from the tracked observer's
view the fused image appears three dimensional. |
Elings |
| |
|
|
 |
Special Feature: Allosphere
Prof. JoAnn Kuchera-Morin, Tom Harper
The UCSB Allosphere is a three-story high spherical immersive
environment, about 30 feet in diameter, with a bridge in the
middle of the instrument, from which up to 25 people will be
able to experience surround-view and surround-sound virtual
worlds in a collaborative fashion. Currently, while we continue
working on the instrumentation of the space, we have about a
quarter to a fifth of the spherical surround-screen lit up. We
will show different demonstrations, ranging from a tour through
one of our colleagues' brains to scientific visualization of
hydrogen bonds, to early renderings of 3D panoramas of locations
on the UCSB campus. |
Elings |
