Search

In this paper we present a novel approach for video shot segmentation based on active vision techniques. A continuity function is introduced taking into account the results from biological vision and experiments and discussions are carried out.

An object-oriented approach to the development of an environment for document processing and analysis is presented. The use of object-oriented techniques may play an important role in the domain we are considering. Document processing systems built today are different from what they were a few years ago: they are larger and more complex. The need of achieving effective and stable systems has gained higher priority. Furthermore, recent systems are expected to devote a deeper attention to the user's perspective. A general model that can be suitably adopted for building such systems is outlined, and it is shown how the object-oriented paradigm can provide a unifying framework for developing the experimental environment. Methodological steps involved are illustrated and a preliminary version of the environment is described. >

For the purpose of enhancing mammographic images, the authors introduce filters based on anisotropic diffusion, and they investigate their relations with classical methods like median, weighted median and tree-structured filters. The performances of the different techniques are quantitatively evaluated and discussed.

In todays hospitals the medical workstation is a basic component of any image management and communication system. The design of such component can be very complex, because of the challenging engineering requirements. In this work we present an architectural model of a flexible and portable software platform upon which medical workstations can be realized. The model is developed within the overall framework provided by the object oriented paradigm. >

Presents a method for the detection of small objects embedded in a noisy background. The detection is performed on the wavelet transformed image. After a preliminary de-noising pass, the objects are separated from background by exploiting the evaluation of Renyi's information at the different decomposition levels of the wavelet transform. We apply the proposed technique to detect microcalcifications in digital mammographic images.

We present results obtained by different contrast enhancement methods applied to medical images. We take into account classical histogram specification, local and wavelet-based techniques and a novel approach for multiscale contrast enhancement. The latter, whose rationale grounds in theories of visual perception, exploits a local definition of the Fechner-Weber's contrast within the context of a non-linear scale-space representation generated by anisotropic diffusion. Our experimental fields concerns a difficult kind of medical images, namely digital mammographic images.

A systematic work has been carried out in order to optimize sample preparation and scanning capacitance microscopy (SCM) on double beveled samples. The method allowed us to enhance depth and lateral resolution and it has been applied to characterize two-dimensional profiles of ultralow energy B implants in Si after diffusion. Implants have been performed into patterned wafers with different stripe widths ranging from 0.5 to 5 microns. B transient enhanced diffusion is strongly reduced with decreasing feature size below about 2 microns. This effect has been related to the high interstitial diffusivity with respect to B. It is even enlarged for lateral diffusion due to the interstitial recombination under the SiO2 mask at the Si/SiO2 interface. The implications for the formation of ultrashallow junctions in device structures are also discussed.

The growing importance of ultra-low energy implantation in Si processing imposes extensive characterization and understanding of such a novel energy regime. In this paper we investigate the evolution of ultra-low energy B implants (0.25–1 keV) after post-implantation annealing, both in terms of atomic diffusion and electrical activation of the doping atoms. Transient enhanced diffusion (TED) of boron after annealing at 900°C is observed even for boron implanted at 250 eV, and also when the implant dose is below the amorphization threshold. At higher temperatures for long anneal times, the TED is overwhelmed by the equilibrium diffusion and it is not visible. However, provided the correct combination of temperatures and times is chosen, the TED can always be observed in samples implanted with a dose at least of 1×10 14 cm 2 . We suggest a possible microscopic mechanism to justify the dependence of the enhanced diffusion of ultra-low energy implanted boron on the implant dose, energy and annealing temperature. An excess of interstitials occurs giving rise to the formation of interstitials-like complexes containing B, probably due to enhanced annihilation of vacancies at the surface. Such an excess of interstitials is able to promote enhanced diffusion of implanted boron, provided the implant dose is high enough to generate a significant total number of point defects. The electrical activation of the ultra-low energy implanted B is shown to be strictly connected to its diffusion. With increasing the dose, the motion of B is supported by the increased amount of ion beam generated interstitials. We have also observed, at the same time, that the electrical activation is favored. The electrical activation, which can be achieved by the ultra-low energy implants we have investigated, is between 10 and 40% after annealing at 1100°C, depending on the implanted dose. For the highest dose we have studied, i.e. 1×10 15 cm 2 , the sheet resistance measured after annealing at 1100°C in a range between 250 eV and 1 keV, is below 1000 Ω. Ultra-low energy implantation is therefore extremely appealing for the future generations of semiconductor devices.

We present a method for the detection of microcalcifications in digital mammographic images. Our approach is based on the wavelet transform, but differently from other techniques proposed in the literature, the detection is directly accomplished in the wavelet domain and no inverse transform is required. After a preliminary denoising pass, microcalcifications are separated from background tissue. This is performed by exploiting information gained through evaluation of Renyi's entropy at the different decomposition levels of the wavelet space. Experimental results achieved on the standard Nimegen data set are shown and discussed.

Segmentation process represents a very difficult task for handwritten recognition aims, due to large variations involved in cursive scripts. This paper describes a method for off-line unconstrained handwritten segmentation, based on a combined use of contour and skeleton derived information. Experiments are discussed and successful segmentation results are reported.

Filters for mammography based on anisotropic diffusion are introduced and evaluated with respect to well known filters adopted within this field. Results reported concern with images including microcalcifications. The performances of anisotropic filters are assessed both on the basis of figures of merit, allowing a preliminary quantitative enhancement evaluation, and according to goal-directed evaluation relying upon a simple segmentation module. Results indicate that proposed filters exhibit an effective and appealing behavior but they are subject to a more complex tuning, if compared to traditional filters.

In this paper we show a new method for efficiently extracting information from contour analysis and demonstrate their use to face the Cursive Script Segmentation challenge. Expressly, our method improves the use of contour to detect the so-called baselines and to use them for a “zone” analysis of the text rows. Furthermore, we show how it is possible to use our method to solve critical situations due to slant and character overhanging. A final discussion about experimental results is also reported.

The paper reports some initial experience in the use of the Actor Model for solving a complex Computer Vision problem: the segmentation of cursive script characters. Our goal is to show how an application programmer can express the different kinds of parallelism present in the problem, emphasizing the advantages obtained by the use of the Actor Model, both from the concurrency and the object-oriented design points of view.

In this paper, we describe a novel Semantic Content Management System (SCMS) able to handle multimedia contents of different kinds (e.g. texts and images) using the related semantics and capable of supporting e-gov applications in different scenarios. All the information is described using semantic metadata semiautomatically extracted from multimedia data, which enriches the browsing experience and enables semantic contents' authoring and queries. To this aim, several Semantic Web technologies have been exploited : RDF/ OWL for data modeling and representation, SPARQL as querying language, Multimedia Information Extraction techniques for content annotation, W3C standard models, vocabularies and micro-formats for resource description. In addition, we propose for entity annotation issue the LOD approach. As an application scenario of the platform, we report a system customization useful for managing the semantic matching between the required professional profiles by a Public Administration and the available skills in a set of curricula vitae with respect to a given call.

Many imaging applications deal with the detection of small targets or spots embedded within an inhomogeneous background. We present a method that accomplishes a multiresolution detection on the wavelettransformed image. The targets are separated from the background by the exploitation of Renyi’s information, which is evaluated at the different decomposition levels of the wavelet transform. The scale-dependent candidate detections are successively combined by means of majority voting for final detection. Connections with results provided in different fields such as multifractal analysis, generalized information measures in scale-space, and cross-entropy analysis in fine-to-coarse transformations are discussed. Detection performance is investigated through an example from medical image analysis. © 2000 Optical Society of America [S0740-3232(00)02107-4] OCIS codes: 100.0100, 100.2000, 100.2960, 100.7410.

Nowadays an ever increasing number of companies rely on Application Service Providers for hosting some or almost all software applications. In this domain, an effective way for customizing user interfaces would be highly desirable. We propose a profiling system interacting with the applications in order to capture users’ behavior. The system is based on Bayesian Networks and we show a strategy for combining data collected in each session and previous knowledge. Experiments on a prototype show that the proposed approach is promising.

The integrated management of multimedia in- formation such as images, graphics, video, audio, and text, is of great interest in a lot of application fields. The major challenges in this non-trivial task are due to structural, syntactic and semantic hetero- geneity of distributed multimedia data repositories. In this paper we present the design of a system, IDEM, based on an object-oriented data model and a wrap- per/mediator architecture, which integrates data from heterogeneous multimedia repositories in a single, unified database view, in order to build a multimedia in- formation system capable of integrating data that re- sides in different data base systems as well as in a variety of non-database data servers.

The explosive growth of the web is at the basis of the great interest into web usage mining techniques in both commercial and research areas. In this paper, a web personalization strategy based on pattern recognition techniques is presented. This strategy takes into account both static information, by means of classical clustering algorithms, and dynamic behavior of a user, proposing a novel and effective re-classification algorithm. Experiments have been carried out in order to validate our approach and evaluate the proposed algorithm.

In this paper we describe the use of NLP techniques and ontologies as the core for building novel e-gov based information systems, and in particular we define the main characteristics of a document management system in the legal domain, that manages a variety of paper documents, automatically transforming them into RDF statements, for suitable indexing, retrieval and long term preservation. Although we describe a general architecture that can be used for several application domains, our system is particularly suitable for the Italian notary realm.

In this paper we describe a prototype system for information retrieval on the Internet. Our idea is that the Web has to be searched both semantically and syntactically. In order to automatically categorize the web pages on the fly we propose a novel approach based on ontology and semantic networks and we describe a prototype system based on the Intelligent Agent Paradigm. Preliminary experiments are shown and discussed while describing open problems and on-going research.

Increasingly, the data and information visualization is becoming strategic for the exploration and explanation of large data sets. The Big Data paradigm pushes for new ways, new technological solutions to deal with the big volume and the big variety of data today. Not surprisingly, a plethora of new tools have emerged, each of them with pros and cons, but all espousing the cause of "Bigness of Data". In this paper, we take one of this emerging tools, namely Neo4J, and stress its capabilities in order to import, query and visualize data coming from a \emph{big} case study: DBpedia. We will describe each step in this study focusing on the used strategies for overcoming the different problems mainly due to the intricate nature of the case study and its volume. We confront with both the intensional schema of DBpedia and its extensional part in order to obtain the best result in its visualization. Finally, an attempt to define some criteria to simplify the large-scale visualization of DBpedia will be made, providing some examples and considerations which have arisen. The ultimate goal of this work is to investigate techniques and approaches to get more insights from the visual representation and analytics of large graph databases.

The definition of ontologies within the multimedia domain still remains a challenging task, due to the complexity of multimedia data and the related knowledge. In this paper, we present a novel framework (MOWIS) aiming at realizing a system for building Multimedia Ontologies from Web Information Sources that has been realized within the PRIN 2007-2009 project Cooperare and presented in previous works. In particular, we propose: i) a multimedia ontology model that combines both low level descriptors and high level semantic concepts; ii) automatic construction of ontologies using the FLICKR web services that provide images, tags, keywords and sometimes useful annotation describing both the image content and personal interesting information. Eventually, we describe an example of automatic ontology generation in a specific domain and present some preliminary experimental results

In the past few years, web usage mining techniques have grown rapidly together with the explosive growth of the web, both in the research and commercial areas. In this work we present a Web mining strategy for Web personalization based on a novel pattern recognition strategy which analyzes and classifies both static and dynamic features. The results of experiments on the data from a large commercial web site are presented to show the effectiveness of the proposed system.

In this work we present a discussion paper from two recent works related to semantic indexing and semantic summarization of documents. In particular, we describe a framework, based on semantic methodologies, that captures the semantic nature of a document written in Natural Language, in order to retrieve a number of semantic RDF triplets and to clusterize these ones using a proper semantic distance. This process can be profitably used both to create an efficient semantic indexing structure for retrieval aims and to automatically generate summaries reflecting the main concepts expressed by the original documents. Several results are also presented and discussed.

The main problem addressed here is the definition of a strategy to improve the sharing and reusability of learning objects in an efficient manner inside an e-Learning community, reducing the cost of creating courses and other learning objects. In particular, in this paper we present a system, for integrating multimedia heterogeneous LO repositories and retrieving distributed data. Such system uses the LOM data model and a wrapper-mediator architecture to perform this task, so providing a simple mechanism for enabling the integration while maintaining the independence of the repositories. Our system also supports a semantic search on managed objects that makes more effective the retrieval process

Context has been defined as the knowledge that can be used to characterize the situation of any entity that is relevant for the (pervasive) system under consideration: given a target application scenario, a context-aware system supports users and devices by providing selective access to the set of data and operations (e.g., interesting services and information, environmental data, close-by people, points of interest etc.) which is relevant in each specific context. More than that, the relative importance of a piece of information to the same user in different contexts or, reciprocally, to different users in the same context may vary enormously; thus the system can personalize information even further by ranking the provided data on the basis of (contextual) user preferences. This chapter presents an introduction to context-aware information management, first providing a literature review and then introducing the main steps needed to design a context-aware system. Context-related problems particularly relevant within pervasive data management are then discussed. We briefly analyze techniques to efficiently associate contexts with information chunks, the evolution issues which arise when the context representation changes over time, the discovery and application of contextual user preferences, and, last but certainly not least, how context awareness can be enforced in the middleware of a pervasive system.