%0 Conference Proceedings
%4 sid.inpe.br/sibgrapi/2013/02.18.16.09
%2 sid.inpe.br/sibgrapi/2013/02.18.16.09.10
%@isbn 978-85-7669-273-7
%T Segmentação/reconstrução de imagens de profundidade associando detecção de bordas e agrupamento
%D 1995
%A Bellon, Olga Regina Pereira,
%A Castanho, José Eduardo Cogo,
%A Tozzi, Clésio Luis,
%@affiliation Departamento de Informática da Universidade Federal do Paraná (UFPR)
%@affiliation Departamento de Engenharia Elétrica da Faculdade de Engenharia e Tecnologia da Universidade Estadual Paulista Júlio de Mesquita Filho (UNESP)
%@affiliation Departamento de Engenharia de Computação e Automação da Faculdade de Engenharia Elétrica da Universidade Estadual de Campinas (UNICAMP)
%E Lotufo, Roberto de Alencar,
%E Mascarenhas, Nelson Delfino d'Ávila,
%B Simpósio Brasileiro de Computação Gráfica e Processamento de Imagens, 8 (SIBGRAPI)
%C São Carlos, SP, Brazil
%8 25-27 Oct. 1995
%I Sociedade Brasileira de Computação
%J Porto Alegre
%P 219-226
%S Anais
%K range image, range image segmentation, reconstruction, clustering, edge detection, plane fitting.
%X In the last years, range sensor technology has been greatly enhanced enabling range images to be more used in computer vision applications. The explicit presence of the range information eliminates the need for processing the image with the objective of extracting 3D information, which is usual with intensity image. Although, other questions remain such as segmentation and reconstruction. The main techniques used to solve the segmentation/reconstruction problem in range image are region growing, split and merge, and clustering. Region growing and split-and-merge algorithms present the problem that they have to deal with threshold values that are difficult to obtain. Clustering algorithms suffer less influence of this kind of problem, although other problems exist. A segmentation/reconstruction technique based on the clustering approach is presented to approximate regions of a range image by planes. The main goal is to reconstruct the image with a minimum error and with simplicity. The association with techniques of range edge detection is proposed to minimize problems of the clustering techniques. Some reconstruction results using simulated images are presented.
%9 Visão por Computador
%@language pt
%3 28 Segmentacao _ reconstrucao.pdf