Skin color-based video segmentation under time-varying illumination

A novel approach for real-time skin segmentation in video sequences is described. The approach enables reliable skin segmentation despite wide variation in illumination during tracking. An explicit second order Markov model is used to predict evolution of the skin-color (HSV) histogram over time. Histograms are dynamically updated based on feedback from the current segmentation and predictions of the Markov model. The evolution of the skin-color distribution at each frame is parameterized by translation, scaling, and rotation in color space. Consequent changes in geometric parameterization of the distribution are propagated by warping and resampling the histogram. The parameters of the discrete-time dynamic Markov model are estimated using Maximum Likelihood Estimation and also evolve over time. The accuracy of the new dynamic skin color segmentation algorithm is compared to that obtained via a static color model. Segmentation accuracy is evaluated using labeled ground-truth video sequences taken from staged experiments and popular movies. An overall increase in segmentation accuracy of up to 24 percent is observed in 17 out of 21 test sequences. In all but one case, the skin-color classification rates for our system were higher, with background classification rates comparable to those of the static segmentation.     

Read-modify-write networks

A read-modify-write register for a set of functions F provides an operation that atomically (1) returns the variables current value v, and (2) replaces that value with f(v), where f is a function in F.A read-modify-write network is a distributed data structure that implements a concurrent, lock-free, low-contention read-modify-write register. For example, counting networks ([3]) are a family of read-modify-write networks that support atomic increments and decrements.We consider the problem of constructing read-modify-write networks for particular sets of functions. Of particular interest are the read-modify-write networks fetch&add, which add to the value an arbitrary number, and fetch&mul, which multiply the value by an arbitrary number.We identify a simple algebraic property of the function set F that requires any distributed read-modify-write implementation for F to have high latency. Any such network has sequential executions in which each token traverses a number of switching elements essentially linear in the number of processes. By contrast, there exist counting networks ([9, 18]) in which tokens traverse a logarithmic number of switches.We provide a matching upper bound for a large class of read-modify-write networks including fetch&add, fetch&mul, and related networks of interest.Received: 15 October 2002, Accepted: 15 May 2003, Published online: 6 February 2004Panagiota Fatourou: Partially supported by funds from the Program E. .E.A.E.K.II of the Greek Ministry of Education. Part of the work of theauthor was performed while at the Max-Planck Institut für Informatik,Saarbrücken, Germany, while at the Department of Computer Science,University of Toronto, Canada, and while visiting the Departmentof Computer Science, Brown University, Providence, USA.Maurice Herlihy: Partially supported by NSF grant 9912401.A preliminary version of this work appears in the Proceedings of the 15th International Symposium on DIStributed Computing, pp. 330-341, Lisbon, Portugal, October 2001. This work also appears as a brief announcement in the Proceedings of the 20th Annual ACM Symposium on Principles of Distributed Computing, Newport, Rhode Island, August 2001.     

Modeling and decoding motor cortical activity using a switching Kalman filter

We present a switching Kalman filter model for the real-time inference of hand kinematics from a population of motor cortical neurons. Firing rates are modeled as a Gaussian mixture where the mean of each Gaussian component is a linear function of hand kinematics. A "hidden state" models the probability of each mixture component and evolves over time in a Markov chain. The model generalizes previous encoding and decoding methods, addresses the non-Gaussian nature of firing rates, and can cope with crudely sorted neural data common in on-line prosthetic applications.     

Interactive volume rendering of thin thread structures within multivalued scientific data sets

We present a threads and halos representation for interactive volume rendering of vector-field structure and describe a number of additional components that combine to create effective visualizations of multivalued 3D scientific data. After filtering linear structures, such as flow lines, into a volume representation, we use a multilayer volume rendering approach to simultaneously display this derived volume along with other data values. We demonstrate the utility of threads and halos in clarifying depth relationships within dense renderings and we present results from two scientific applications: visualization of second-order tensor valued magnetic resonance imaging (MRI) data and simulated 3D fluid flow data. In both application areas, the interactivity of the visualizations proved to be important to the domain scientists. Finally, we describe a PC-based implementation of our framework along with domain specific transfer functions, including an exploratory data culling tool, that enable fast data exploration.     

Learning the Statistics of People in Images and Video

This paper address the problems of modeling the appearance of humans and distinguishing human appearance from the appearance of general scenes. We seek a model of appearance and motion that is generic in that it accounts for the ways in which people's appearance varies and, at the same time, is specific enough to be useful for tracking people in natural scenes. Given a 3D model of the person projected into an image we model the likelihood of observing various image cues conditioned on the predicted locations and orientations of the limbs. These cues are taken to be steered filter responses corresponding to edges, ridges, and motion-compensated temporal differences. Motivated by work on the statistics of natural scenes, the statistics of these filter responses for human limbs are learned from training images containing hand-labeled limb regions. Similarly, the statistics of the filter responses in general scenes are learned to define a background distribution. The likelihood of observing a scene given a predicted pose of a person is computed, for each limb, using the likelihood ratio between the learned foreground (person) and background distributions. Adopting a Bayesian formulation allows cues to be combined in a principled way. Furthermore, the use of learned distributions obviates the need for hand-tuned image noise models and thresholds. The paper provides a detailed analysis of the statistics of how people appear in scenes and provides a connection between work on natural image statistics and the Bayesian tracking of people.     

A Model of Partially Observable State Game and its Optimality

In this paper we present a model of two-players partially observable state-game and study its optimality. The model is inspired by the practical problem of negotiation in a multi-agent system and formulates, from a the game theory point of view, the so-called contract net protocol. It covers a wide variety of real problems including some simple card games such as blackjack, and many negotiation and bargaining situations. The results that follow are valid for non-zero-sum games as well as for zero-sum games. Basically, we establish and prove the relation between partially observable state games and some classical (single-state) bi-matrix games. If the original state game is zero-sum, then the equivalent bi-matrix game is so.     

Distance estimation and object location via rings of neighbors

We consider four problems on distance estimation and object location which share the common flavor of capturing global information via informative node labels: low-stretch routing schemes [48], distance labeling [25], searchable small worlds [31], and triangulation-based distance estimation [34]. Focusing on metrics of low doubling dimension, we approach these problems with a common technique called rings of neighbors, which refers to a sparse distributed data structure that underlies all our constructions. Apart from improving the previously known bounds for these problems, our contributions include extending Kleinberg’s small world model to doubling metrics, and a short proof of the main result in Chan et al. [15]. Doubling dimension is a notion of dimensionality for general metrics that has recently become a useful algorithmic concept in the theoretical computer science literature. This work was done when A. Slivkins was a graduate student at Cornell University and was supported by the Packard Fellowship of Jon Kleinberg. Preliminary version of this paper has appeared in 24th Annual ACM SIGACT-SIGOPS Symposium on Principles of Distributed Computing (PODC), 2005.     

A curve fitting procedure to derive inter-annual phenologies from time series of noisy satellite NDVI data

Annual, inter-annual and long-term trends in time series derived from remote sensing can be used to distinguish between natural land cover variability and land cover change. However, the utility of using NDVI-derived phenology to detect change is often limited by poor quality data resulting from atmospheric and other effects. Here, we present a curve fitting methodology useful for time series of remotely sensed data that is minimally affected by atmospheric and sensor effects and requires neither spatial nor temporal averaging. A two-step technique is employed: first, a harmonic approach models the average annual phenology; second, a spline-based approach models inter-annual phenology. The principal attributes of the time series (e.g., amplitude, timing of onset of greenness, intrinsic smoothness or roughness) are captured while the effects of data drop-outs and gaps are minimized. A recursive, least squares approach captures the upper envelope of NDVI values by upweighting data values above an average annual curve. We test this methodology on several land cover types in the western U.S., and find that onset of greenness in an average year varied by less than 8 days within land cover types, indicating that the curve fit is consistent within similar systems. Between 1990 and 2002, temporal variability in onset of greenness was between 17 and 35 days depending on the land cover type, indicating that the inter-annual curve fit captures substantial inter-annual variability. Employing this curve fitting procedure enhances our ability to measure inter-annual phenology and could lead to better understanding of local and regional land cover trends.     

Hypergossiping: A generalized broadcast strategy for mobile ad hoc networks

Broadcasting is a commonly used communication primitive needed by many applications and protocols in mobile ad hoc networks (MANET). Unfortunately, most broadcast solutions are tailored to one class of MANETs with respect to node density and node mobility and are unlikely to operate well in other classes. In this paper, we introduce hypergossiping, a novel adaptive broadcast algorithm that combines two strategies. Hypergossiping uses adaptive gossiping to efficiently distribute messages within single network partitions and implements an efficient heuristic to distribute them across partitions. Simulation results in ns-2 show that hypergossiping operates well for a broad range of MANETs with respect to node densities, mobility levels and network loads.