Submitted by Elizabete Silva (elizabete.silva@ufes.br) on 2015-09-02T19:11:11Z No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Utilizando Contexto na Representa¸c˜ao de Imagens para a.pdf: 4083803 bytes, checksum: e0ced4975f7eee5db5316f7e096db639 (MD5) / Approved for entry into archive by Morgana Andrade (morgana.andrade@ufes.br) on 2015-11-23T19:25:43Z (GMT) No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Utilizando Contexto na Representa¸c˜ao de Imagens para a.pdf: 4083803 bytes, checksum: e0ced4975f7eee5db5316f7e096db639 (MD5) / Made available in DSpace on 2015-11-23T19:25:43Z (GMT). No. of bitstreams: 2 license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Utilizando Contexto na Representa¸c˜ao de Imagens para a.pdf: 4083803 bytes, checksum: e0ced4975f7eee5db5316f7e096db639 (MD5) Previous issue date: 2014 / A classifica¸c˜ao de cenas ´e um campo bastante popular na ´area de vis˜ao computacional e encontra diversas aplica¸c˜oes tais como: organiza¸c˜ao e recupera¸c˜ao de imagem baseada em conte´udo, localiza¸c˜ao e navega¸c˜ao de robˆos. No entanto, a classifica¸c˜ao autom´atica de cenas ´e uma tarefa desafiadora devido a diversos fatores, tais como, ocorrˆencia de oclus˜ao, sombras, reflex˜oes e varia¸c˜oes nas condi¸c˜oes de ilumina¸c˜ao e escala. Dentre os trabalhos que objetivam solucionar o problema da classifica¸c˜ao autom´atica de cenas, est˜ao aqueles que utilizam transformadas n˜ao-param´etricas e aqueles que tˆem obtido melhora no desempenho de classifica¸c˜ao atrav´es da explora¸c˜ao da informa¸c˜ao contextual. Desse modo, esse trabalho prop˜oe dois descritores de imagens que associam informa¸c˜ao contextual, ou seja, informa¸c˜ao advinda de regi˜oes vizinhas, a um tipo de transformada n˜ao-param´etrica. O objetivo ´e propor uma abordagem que n˜ao eleve demasiadamente a dimens˜ao do vetor de caracter´ısticas e que n˜ao utilize a t´ecnica de representa¸c˜ao intermedi´aria bag-of-features, diminuindo, assim, o custo computacional e extinguindo a necessidade de informa¸c˜ao de parˆametros, o que possibilita a sua utiliza¸c˜ao por usu´arios que n˜ao possuem conhecimento na ´area de reconhecimento de padr˜oes. Assim, s˜ao propostos os descritores CMCT (Transformada Census Modificada Contextual) e ECMCT (CMCT Estendido) e seus desempenhos s˜ao avaliados em quatro bases de dados p´ublicas. S˜ao propostas tamb´em cinco varia¸c˜oes destes descritores (GistCMCT, GECMCT, GistCMCT-SM, ECMCT-SM e GECMCT-SM), obtidas atrav´es da associa¸c˜ao de cada um deles com outros descritores. Os resultados obtidos nas quatro bases de dados mostram que as representa¸c˜oes propostas s˜ao competitivas, e que provocam um aumento nas taxas de classifica¸c˜ao, quando comparados com outros descritores. / Scene classification is a very popular topic in the field of computer vision and it has many applications, such as, content-based image organization and retrieval and robot navigation. However, scene classification is quite a challenging task, due to the occurrence of occlusion, shadows and reflections, illumination changes and scale variability. Among the approaches to solve the scene classification problems are those that use nonparametric transform and those that improve classification results by using contextual information. Thus, this work proposes two image descriptors that associate contextual information, from neighboring regions, with a non-parametric transforms. The aim is to propose an approach that does not increase excessively the feature vector dimension and that does not use the bag-of-feature method. In this way, the proposals descrease the computational costs and eliminate the dependence parameters, which allows the use of those descriptors in applications for non-experts in the pattern recognition field. The CMCT and ECMCT descriptors are presented and their performances are evaluated, using four public datasets. Five variations of those descriptors are also proposed (GistCMCT, GECMCT, GistCMCT-SM, ECMCT-SM e GECMCT-SM), obtained through their association with other approaches. The results achieved on four public datasets show that the proposed image representations are competitive and lead to an increase in the classification rates when compared to others descriptors.