The MobileASL project aims to increase accessibility by enabling Deaf people to communicate over video cell phones in their native language, American Sign Language (ASL). Real-time video over cell phones can be a computationally intensive task that quickly drains the battery, rendering the cell phone useless. Properties of conversational sign language allow us to save power and bits: namely, lower frame rates are possible when one person is not signing due to turntaking, and signing can potentially employ a lower frame rate than fingerspelling. We conduct a user study with native signers to examine the intelligibility of varying the frame rate based on activity in the video. We then describe several methods for automatically determining the activity of signing or not signing from the video stream in real-time. Our results show that varying the frame rate during turn-taking is a good way to save power without sacrificing intelligibility, and that automatic activity analysis is feasible.

Background subtraction is a crucial step in many automatic video content analysis applications. While numerous acceptable techniques have been proposed so far for background extraction, there is still a need to produce more efficient algorithms in terms of adaptability to multiple environments, noise resilience, and computation efficiency. In this paper, we present a powerful method for background extraction that improves in accuracy and reduces the computational load. The main innovation concerns the use of a random policy to select values to build a samples-based estimation of the background. To our knowledge, it is the first time that a random aggregation is used in the field of background extraction. In addition we propose a novel policy that propagates information between neighboring pixels of an image. Experiment detailed in this paper show how our method improves on other widely used techniques, and how it outperforms these techniques for noisy images.

vehicles at urban road intersections

Multi-temporal classification of remote sensing images is a challenging problem, in which efficient combination of different sources of information (e.g. temporal, contextual, or multi-sensor) can improve the results. In this paper, we present a general framework based on kernel methods for the integration of heterogeneous sources of information. Using the theoretical principles in this framework, three main contributions are presented. First, a novel family of kernel-based methods for multi-temporal classification of remote sensing images is presented. The second contribution is the development of non-linear kernel classifiers for the well-known difference and ratioing change detection methods, by formulating them in an adequate high dimensional feature space. Finally, the presented methodology allows the integration of contextual information and multi-sensor images with different levels of non-linear sophistication. The binary support vector classifier (SVC) and the one-class support vector domain

For the anomalous change detection problem, you have a pair of images, taken of the same scene, but at different times and typically under different viewing conditions. You are looking for interesting differences between the two images. There will be some differences that are pervasive, perhaps due to overall contrast, brightness or focus differences, or maybe due to atmospheric or even seasonal changes – but there may also be changes that occur in only a few pixels. These rare changes are potentially indicative of something truly changing in the scene, and the idea is to use anomaly detection to find them. But you want to identify the changes that are unusual. You do not want to be confounded by ususual pixels that are “similarly unusual ” in both images. We propose a machine learning framework for identifying these anomalous changes. 1.

Abstract—In this work we consider an event-driven Wireless Visual Sensor Network (WVSN) where each camera node transmits a frame to the cluster-head only if an event of interest was captured in the frame for energy and bandwidth conservation. Specifically, we consider the scenario where each camera node receives decision support from an independent but possibly attacked (and hence error-prone) scalar-sensor regarding the presence or absence of an event. We study the overall detection performance achieved by various techniques that utilize the scalar and image-based decisions. We conclude that in image sequences involving extraneous lighting and background changes (such as in the case of outdoor surveillance), the combination techniques generally achieve a lower total probability of error. I.

Change detection between two images is challenging and needed in a wide variety of imaging applications. Several approaches have been yet developed, especially methods based on difference image. In this paper, we propose an original patch-based Markov modeling framework to detect spatial irregularities in the difference image with low false alarm rates. Experimental results show that the proposed approach performs well for change detection, especially for images with low signal-to-noise ratios. 1.

Automated video surveillance applications require accurate separation of foreground and background image content. Cost sensitive embedded platforms place realtime performance and efficiency demands on techniques to accomplish this task. In this paper we evaluate pixel-level foreground extraction techniques for a low cost integrated surveillance system. We introduce a new adaptive technique, multimodal mean (MM), which balances accuracy, performance, and efficiency to meet embedded system requirements. Our evaluation compares several pixel-level foreground extraction techniques in terms of their computation and storage requirements, and functional accuracy for three representative video sequences. The proposed MM algorithm delivers comparable accuracy of the best alternative (Mixture of

Abstract—In this work we consider an event-driven wireless visual sensor network (WVSN) comprised of untethered camera nodes and scalar sensors deployed in a hostile environment. In the event-driven paradigm, each camera node transmits a surveillance frame to the cluster-head only if an event of interest was captured in the frame, for energy and bandwidth conservation. We thus examine a simple image processing algorithm at the camera nodes based on difference frames and the chi-squared detector. We show that the test statistic of the chi-squared detector is equivalent to that of a robust (non-parametric) detector and that this simple algorithm performs well on indoor surveillance sequences and some, but not all, outdoor sequences. In outdoor sequences containing significant changes in background and lighting, this simple detector may produce a high probability of error and benefits from the inclusion of scalar sensor decisions. The scalar sensor decisions are, however, prone to attack and may exhibit errors that are arbitrarily frequent, pervasive throughout the network and difficult to predict. To achieve attack prediction and mitigation given an attacker whose actions are not known a priori, we employ game-theoretic analysis. We show that the scalar sensor error can be controlled through cluster-head checking and appropriate selection of cluster size. Given this attack mitigation, we employ real-life sequences to determine the total probability of error when individual and combined decisions are utilized and we discuss the ensuing ramifications and performance issues. Index Terms—Actuation, event-detection, game theory, scalarsensors, sensor network security, wireless visual sensor networks

www.vision.deis.unibo.it Abstract. Time-of-Flight (TOF) cameras relate to a very recent and growing technology which has already proved to be useful for computer vision tasks. In this paper we investigate on the use of a TOF camera to perform video-based graffiti detection, which can be thought of as a monitoring system able to detect acts of vandalism such as dirtying, etching and defacing walls and objects surfaces. Experimental results show promising capabilities of the proposed approach, with improvements expected as the technology gets more mature. 1