Close
Metadata

@MastersThesis{Mota:1984:MéReAm,
               author = "Mota, Fernando Antonio Araujo",
                title = "Um m{\'e}todo para redu{\c{c}}{\~a}o de ambiguidade em 
                         segmenta{\c{c}}{\~a}o de imagens",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "1984",
              address = "S{\~a}o Jos{\'e} dos Campos",
                month = "1984-08-10",
             keywords = "computer vision, scene analysis, relaxation operation.",
             abstract = "No desenvolvimento de sistemas mecanizados de 
                         percep{\c{c}}{\~a}o visual distingue-se, atualmente, um conjunto 
                         de fases que caracterizam a modelagem computacional do processo 
                         perceptivo. Nas fases de segmenta{\c{c}}{\~a}o e interpreta 
                         al{\'e}m do compromisso a respeito da elimina{\c{c}}{\~a}o de 
                         in consiste depara-se com o problema da ambiguidade. Prop{\~o}e 
                         neste trabalho uma abordagem para tratar, ao n- de identifica de 
                         objeto, tais quest{\~o}es atrav{\'e}s de relacionamentos sema e 
                         de opera{\c{c}}{\~o}es paralelas iteradas (opera{\c{c}}{\~o}es 
                         de relaxa{\c{c}}{\~a}o) sobre um modelo discreto de 
                         rotula{\c{c}}{\~a}o de segmentos (grafos rotulados). Pretende-se 
                         reduzir ao m{\'a}ximo a ambiguidade atrav{\'e}s de um {\'u}nico 
                         processo de elimina{\c{c}}{\~a}o sucessiva, em lugar de um 
                         processo para cada segmenta{\c{c}}{\~a}o poss{\'{\i}}vel. A 
                         disserta{\c{c}}{\~a}o fundamenta-se, principalmente, na 
                         caracteriza{\c{c}}{\~a}o alg{\'e}brica do operador de 
                         relaxa{\c{c}}{\~a}o e na argumenta{\c{c}}{\~a}o formal, 
                         atrav{\'e}s da {\'a}lgebra discreta, que evidencia a sua 
                         validade computacional. ABSTRACT: Nowadays, in the development of 
                         automatic systems of visual perception, a set of phases that 
                         characterize the computational modelling of the perceptive process 
                         are distinguished. During the segmentation and interpretation 
                         phases, besides the compromise regarding the elimination of 
                         inconsistencies, there appears the ambiguity problem. In this work 
                         it is proposed an approach to deal with those questions, in the 
                         level of the identification of the object, using the semantic 
                         relationships and iterated parallel operations (relaxation 
                         operations) upon a discrete model that labels segments (labelled 
                         graphs). It is intended to reduce to the maximum the ambiguity 
                         through a s-ingle process of successive elimination, instead of a 
                         process for each possible segmentation. The dissertation is based 
                         mainly in the algebraic characterization of the relaxation 
                         operator and in the formal treatment, through discrete algebra, 
                         which puts into evidence its computational validity.",
            committee = "Souza, Celso de Renna e (presidente) and Velasco, Fl{\'a}vio 
                         Roberto Dias (orientador) and Sakane, Fernando Toshinori and 
                         Mascarenhas, Nelson Delfino d´{\'A}vila and Simoni, Paulo 
                         Ouvera",
           copyholder = "SID/SCD",
         englishtitle = "x",
             language = "pt",
                pages = "118",
                  ibi = "8JMKD3MGP8W/352RL5L",
                  url = "http://urlib.net/rep/8JMKD3MGP8W/352RL5L",
           targetfile = "publicacao.pdf",
        urlaccessdate = "2021, Apr. 18"
}


Close