• Login
    View Item 
    •   DSpace Home
    • BIBLIOTECA DIGITAL: Teses & Dissertações
    • Teses & Dissertações
    • View Item
    •   DSpace Home
    • BIBLIOTECA DIGITAL: Teses & Dissertações
    • Teses & Dissertações
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Segmentaçao de imagens baseada em dependencia espacial utilizando campo aleatório de Markov associado com características de texturas

    Thumbnail
    View/Open
    Disserta.pdf (1.639Mb)
    Date
    2005
    Author
    Schwartz, William Robson
    Metadata
    Show full item record
    Subject
    Teses
    Sistemas de reconhecimento de padrões
    Processamentos de imagens - Termicas digitais
    Ciencia da Computação
    xmlui.dri2xhtml.METS-1.0.item-type
    Dissertação
    Abstract
    Resumo: Uma etapa crítica presente no processo de análise de imagens é a segmentação, responsável por obter informações de alto n'nível sobre as regiões ou objetos contidos na imagem, de modo a facilitar sua interpretação. Contudo, a segmentação ainda é um dos maiores desafios na área de análise de imagens, particularmente quando não se utiliza informações previamente adquiridas sobre a imagem a ser segmentada. Os métodos convencionais de segmentação desconsideram a dependência espacial entre as regiões, o que pode gerar resultados impróprios. Técnicas que consideram a dependência espacial entre as regiões da imagem têm recebido crescente atenção da comunidade científica, pois apresentam uma maior precisão nos resultados obtidos. Embora avanços significativos tenham sido alcançados na segmentação de texturas e de imagens coloridas separadamente, a combinação dessas duas propriedades é considerada como um problema bem mais complexo. Devido a importância dessa etapa no processo de análise de imagens e ao fato de não existirem soluções definitivas para o problema, este trabalho propõe o desenvolvimento de um novo método de segmentação aplicado a imagens texturizadas monocromáticas e coloridas. O método utiliza a formulação Bayesiana para associar a dependência espacial modelada por um campo aleatório de Markov com características de texturas. A segmentação final é obtida por meio da aplicação de t'cênicas de relaxação para minimizar uma função de energia definida a partir da referida associação. Experimentos são efetuados visando avaliar os métodos de análise de texturas, bem como validar a metodologia proposta.
     
    Abstract: A critical stage present in the image analysis process is the segmentation, responsible for obtaining high level information about regions or objects in the image, in order to facilitate its interpretation. However, the segmentation is still one of the greatest challenges in the image analysis area, particularly when it does not use information previously acquired on the image to be segmented. Conventional segmentation methods do not consider the spatial dependence between the regions, which can generate improper results. Techniques considering the spatial dependence between the image regions have received increasing attention from the scientific community, because they present a major precision in the obtained results. Although significant advances have been reached in the segmentation of textures and colored images separately, the combination of these two properties is considered a more complex problem. Due to the importance of this stage in the image analysis process and to the fact that does not exist definitive solutions to the problem, this work considers the development of a new segmentation method applied to gray scale and color texture images. The method uses the Bayesian formulation to associate the spatial dependence modeled by a Markov random field with texture features. The final segmentation is obtained by the application of relaxation techniques to minimize an energy function defined by such association. Experiments are performed to evaluate the texture analysis methods, as well as validating the proposal method.
     
    URI
    https://hdl.handle.net/1884/1597
    Collections
    • Teses & Dissertações [6988]

    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV
     

     

    Browse

    All of DSpaceCommunities & CollectionsBy Issue DateAuthorsTitlesSubjectsxmlui.ArtifactBrowser.Navigation.browse_typeThis CollectionBy Issue DateAuthorsTitlesSubjectsxmlui.ArtifactBrowser.Navigation.browse_type

    My Account

    LoginRegister

    Statistics

    View Usage Statistics

    DSpace software copyright © 2002-2016  DuraSpace
    Contact Us | Send Feedback
    Theme by 
    Atmire NV