We have developed a near-real-time computer system that can locate and track subject's head, then recognize the person by comparing characteristics of face to those known individuals. The computational approach taken in this is motivated both physiology information theory, as well practical requirements performance accuracy. Our treats recognition problem an intrinsically two-dimensional (2-D) rather than requiring recovery three-dimensional geometry, taking advantage fact faces are normally...

An approach to the detection and identification of human faces is presented, a working, near-real-time face recognition system which tracks subject's head then recognizes person by comparing characteristics those known individuals described. This treats as two-dimensional problem, taking advantage fact that are normally upright thus may be described small set 2-D characteristic views. Face images projected onto feature space ('face space') best encodes variation among images. The defined...

Vision-based hand gesture interfaces require fast and extremely robust detection. Here, we study view-specific posture detection with an object recognition method proposed by Viola Jones. Training this is computationally very expensive, prohibiting the evaluation of many appearances for their suitability to In paper, present a frequency analysis-based instantaneous estimation class separability, without need any training. We built detectors most promising candidates, receiver operating...

Computer-mediated communication systems known as collaborative virtual environments (CVEs) allow geographically separated individuals to interact verbally and nonverbally in a shared space real time. We discuss CVE-based research paradigm that transforms (i.e., filters modifies) nonverbal behaviors during social interaction. Because the technology underlying CVEs allows strategic decoupling of rendered behavior from actual interactants, conceptual perceptual constraints inherent face-to-face...

The performance of face recognition methods using subspace projection is directly related to the characteristics their basis images, especially in cases local distortion or partial occlusion. In order for a method be robust and occlusion, images generated by should exhibit part-based representation. We propose an effective representation named locally salient ICA (LS-ICA) that LS-ICA only employs information from important facial parts maximize benefit applying idea "recognition parts". It...

We present a non-photorealistic rendering approach to capture and convey shape features of real-world scenes. use camera with multiple flashes that are strategically positioned cast shadows along depth discontinuities in the scene. The projective-geometric relationship camera-flash setup is then exploited detect distinguish them from intensity edges due material discontinuities.We introduce depiction methods utilize detected edge generate stylized static animated images. can highlight...

We propose a state-based approach to gesture learning and recognition. Using spatial clustering temporal alignment, each is defined be an ordered sequence of states in spatial-temporal space. The 2D image positions the centers head both hands user are used as features; these located by color-based tracking method. From training data given gesture, we first learn information then group into segments that automatically aligned temporally. further integrated build finite state machine (FSM)...

This paper introduces "Flocks of Features," a fast tracking method for non-rigid and highly articulated objects such as hands. It combines KLT features learned foreground color distribution to facilitate 2D position from monocular view. The tracker's benefits lie in its speed, robustness against background noise, ability track that undergo arbitrary rotations vast rapid deformations. We demonstrate tracker performance on hand with non-stationary camera unconstrained indoor outdoor...

We describe a framework and prototype implementation for unobtrusive mobile remote collaboration on tasks that involve the physical environment. Our system uses Augmented Reality paradigm model-free, markerless visual tracking to facilitate decoupled, live updated views of environment world-stabilized annotations while supporting moving camera unknown, unprepared environments. In order evaluate our concept prototype, we conducted user study with 48 participants in which expert instructed...

Immersive collaborative virtual environments (CVEs) are simulations in which geographically separated individuals interact a shared, three-dimensional, digital space using immersive environment technology. Unlike videoconference technology, transmits direct video streams, CVEs accurately track movements of interactants and render them nearly simultaneously (i.e., real time) onto avatars, three-dimensional representations the interactants. Nonverbal behaviors can be rendered veridically or...

Advances in imaging technology have made the capture and display of digital images ubiquitous. A variety displays are used to view them, ranging from high-resolution computer monitors low-resolution mobile devices, often undergo changes size aspect ratio adapt different screens. Also, displaying printing documents with embedded frequently entail resizing comply overall layout. Straightforward image operators, such as scaling, do not produce satisfactory results, since they oblivious content....

We present a mobile augmented reality (AR) translation system, using smartphone's camera and touchscreen, that requires the user to simply tap on word of interest once in order produce translation, presented as an AR overlay. The seamlessly replaces original text live stream, matching background foreground colors estimated from source images. For this purpose, we developed efficient algorithm for accurately detecting location orientation stream is robust perspective distortion, combine it...

Algorithms based on RANSAC that estimate models using feature correspondences between images can slow down tremendously when the percentage of correct (inliers) is small. In this paper, we present a probabilistic parametric model allows us to assign confidence values for each matching correspondence and therefore accelerates generation hypothesis under these conditions. Our framework leverages Extreme Value Theory accurately statistics scores produced by nearest-neighbor matcher. Using new...

The motivations for organizing this special section were to better address the challenges of face recognition in real-world scenarios, promote systematic research and evaluation promising methods systems, provide a snapshot where we are domain, stimulate discussion about future directions. We solicited original contributions on all aspects recognition, including: design robust similarity features metrics; clustering sorting algorithms; novel user interaction models algorithms tagging;...

The computational capability of mobile phones has been rapidly increasing, to the point where augmented reality become feasible on cell phones. We present an approach indoor localization and pose estimation in order support applications a phone platform. Using embedded camera, application localizes device familiar environment determines its orientation. Once 6 DOF is determined, 3D virtual objects from database can be projected into image displayed for user. Off-line data acquisition...

In this paper we present a user study evaluating the benefits of geometrically correct user-perspective rendering using an Augmented Reality (AR) magic lens. simulation compared lens against common device-perspective on both phone-sized and tablet-sized displays. Our results indicate that display allows for significantly faster performance selection task has over task. Based these promising results, created proof-of-concept prototype, engineered with current off-the-shelf devices software....

We propose a novel formulation of minimal case solutions for determining the relative pose perspective and generalized cameras given partially known rotation, namely, axis rotation. An rotation may be easily obtained by detecting vertical vanishing points with computer vision techniques, or aid sensor measurements from smart phone. Given our algorithms solve angle around along unknown translation. formulate these problems as Quadratic Eigen value Problems which are very simple to construct....

Producing sports highlights is a labor-intensive work that requires some degree of specialization. We propose model capable automatically generating with focus on cricket. Cricket sport complex set rules and played for longer time than most other sports. In this paper we considers both event-based excitement-based features to recognize clip important events in cricket match. Replays, audio intensity, player celebration, playfield scenarios are examples cues used capture such events. To...

A simple imaging range sensor is described, based on the measurement of focal error, as described by A. Pentland (1982 and 1987). The current implementation can produce over a 1 m/sup 3/ workspace with measured standard error 2.5% (4.5 significant bits data). system implemented using relatively inexpensive commercial image-processing equipment. Experience shows that this ranging technique be both economical practical for tasks which require quick reliable but coarse estimates range. Examples...

The research described in this paper analyzes the in-plane rotational robustness of Viola-Jones object detection method when used for hand appearance detection. We determine bounds training and achieving undiminished performance without an increase classifier complexity. result - up to 15/spl deg/ total differs from method's on faces (30/spl total). found that randomly rotating data within these allows rates about one order magnitude better than those trained strictly aligned data....

We propose a novel approach for modeling, tracking and recognizing facial expressions. Our method works on low dimensional expression manifold, which is obtained by Isomap embedding. In this space, contour features are first clustered, using mixture model. Then, dynamics learned classification. use ICondensation to track in the embedded while expressions cooperative manner, within common probabilistic framework. The image observation likelihood derived from variation of Active Shape Model...