On restricted one-counter machines

Let be the class of real-time nondeterministic one-counter machines whose counters make at mostone reversal. Let ₁ (respectively, ₂) be the subclass consisting of machines whose only nondeterministic move is in the choice of when to reverse the counter (respectively, when to start using the counter). ₁ and ₂ are among the simplest known classes of machines for which the universe problem has been shown undecidable. (The universe problem for a class of machines is the problem of deciding if an arbitrary machine in the class accepts all its inputs.) Here, we show that the classes of languages accepted by machines in ₁ and ₂ are incomparable. Moreover, the union of the language classes is properly contained in the class defined by . We also, briefly, look at the closure properties of these machines.This research was supported in part by NSF Grant MCS78-01736.     

Inflating examples to obtain rules

A new machine learning system is presented in this article. It is called INNER and induces classification rules from a set of training examples. The process followed by this system starts with the random selection of a subset of examples that are iteratively inflated in order to cover the surroundings provided that they are inhabited by examples of the same class, thus becoming rules that will be applied by means of a partial matching mechanism. The rules so obtained can be seen as clusters of examples and represent clear evidence to support explanations about their future classifications and may be used to build intelligent advisors. The whole algorithm can be seen as a set of elastic transformations of examples and rules and produces concise, accurate rule sets, as is experimentally demonstrated in the final section of the article. © 2003 Wiley Periodicals, Inc.     

An iterative thresholding algorithm for image segmentation

A thresholding technique is developed for segmenting digital images with bimodal reflectance distributions under nonuniform illumination. The algorithm works in a raster format, thus making it an attractive segmentation tool in situations requiring fast data throughput. The theoretical base of the algorithm is a recursive Taylor expansion of a continuously varying threshold tracking function.     

2‐D/3‐D switchable displays

Abstract— In this paper, the design of a lenticular‐based 2‐D/3‐D display for mobile applications is described. This display combines look‐around capability with good 3‐D resolution. In order to allow high‐resolution datagraphic applications, a concept based on actively switched lenses has been developed. A very noticeable problem for such displays is the occurrence of dark bands. Despite slanting the lenticular and defocusing the lens, banding becomes unacceptable when the display is viewed from an angle. As a solution, fractional viewing systems to reduce the banding intensity by almost two orders of magnitude is introduced. The resulting 3‐D display can be viewed from any horizontal direction without banding.     

Distributed algorithms for barrier coverage using relocatable sensors

We study the barrier coverage problem using relocatable sensor nodes. We assume each sensor can sense an intruder or event inside its sensing range. Sensors are initially located at arbitrary positions on the barrier and can move along the barrier. The goal is to find final positions for sensors so that the entire barrier is covered. In recent years, the problem has been studied extensively in the centralized setting. In this paper, we study a barrier coverage problem in the distributed and discrete setting. We assume that we have n identical sensors located at grid positions on the barrier, and that each sensor repeatedly executes a Look-Compute-Move cycle: based on what it sees in its vicinity, it makes a decision on where to move, and moves to its next position. We make two strong but realistic restrictions on the capabilities of sensors: they have a constant visibility range and can move only a constant distance in every cycle. In this model, we give the first two distributed algorithms that achieve barrier coverage for a line segment barrier when there are enough nodes in the network to cover the entire barrier. Our algorithms are synchronous, and local in the sense that sensors make their decisions independently based only on what they see within their constant visibility range. One of our algorithms is oblivious whereas the other uses two bits of memory at each sensor to store the type of move made in the previous step. We show that our oblivious algorithm terminates within \(\varTheta (n^2)\) steps with the barrier fully covered, while the constant-memory algorithm is shown to take \(\varTheta (n)\) steps to terminate in the worst case. Since any algorithm in which a sensor can only move a constant distance in one step requires \(\varOmega (n)\) steps on some inputs, our second algorithm is asymptotically optimal.     

Link-Node Characterization for Energy-Efficient Communication in a Static Random Wireless Sensor Network

Studies of wireless networks often work with a relatively high communication range so that each node typically communicates with a high number of neighbors. This is necessary, with a random placement, to avoid loss of coverage or connectivity. However, communicating through remote neighbors requires more energy, which can be a severe disadvantage since energy is often critical in this type of network. By comparing two-ranges, one to ensure coverage and one to save energy, we identify link-nodes which play an essential role to keep the network connected. Characterizing these nodes and their occurrence in communication routes will help to save energy and increase lifetime of the overall network. Without precautions, these link-nodes receive a high traffic load. Using simulations we find that slightly longer routes can avoid link-nodes and thus reduce average energy consumption. Network lifetime can be extended by identifying these nodes at startup and reducing their energy consumption wherever possible.     

Sonar image segmentation using an unsupervised hierarchical MRFmodel

This paper is concerned with hierarchical Markov random field (MRP) models and their application to sonar image segmentation. We present an original hierarchical segmentation procedure devoted to images given by a high-resolution sonar. The sonar image is segmented into two kinds of regions: shadow (corresponding to a lack of acoustic reverberation behind each object lying on the sea-bed) and sea-bottom reverberation. The proposed unsupervised scheme takes into account the variety of the laws in the distribution mixture of a sonar image, and it estimates both the parameters of noise distributions and the parameters of the Markovian prior. For the estimation step, we use an iterative technique which combines a maximum likelihood approach (for noise model parameters) with a least-squares method (for MRF-based prior). In order to model more precisely the local and global characteristics of image content at different scales, we introduce a hierarchical model involving a pyramidal label field. It combines coarse-to-fine causal interactions with a spatial neighborhood structure. This new method of segmentation, called the scale causal multigrid (SCM) algorithm, has been successfully applied to real sonar images and seems to be well suited to the segmentation of very noisy images. The experiments reported in this paper demonstrate that the discussed method performs better than other hierarchical schemes for sonar image segmentation.     

Generalized overlap measures for evaluation and validation in medical image analysis

Measures of overlap of labelled regions of images, such as the Dice and Tanimoto coefficients, have been extensively used to evaluate image registration and segmentation algorithms. Modern studies can include multiple labels defined on multiple images yet most evaluation schemes report one overlap per labelled region, simply averaged over multiple images. In this paper, common overlap measures are generalized to measure the total overlap of ensembles of labels defined on multiple test images and account for fractional labels using fuzzy set theory. This framework allows a single "figure-of-merit" to be reported which summarises the results of a complex experiment by image pair, by label or overall. A complementary measure of error, the overlap distance, is defined which captures the spatial extent of the nonoverlapping part and is related to the Hausdorff distance computed on grey level images. The generalized overlap measures are validated on synthetic images for which the overlap can be computed analytically and used as similarity measures in nonrigid registration of three-dimensional magnetic resonance imaging (MRI) brain images. Finally, a pragmatic segmentation ground truth is constructed by registering a magnetic resonance atlas brain to 20 individual scans, and used with the overlap measures to evaluate publicly available brain segmentation algorithms.