Filtragem MAP 2-D de imagens CT ruidosas

Detalhes bibliográficos
Ano de defesa: 2011
Autor(a) principal: Geraldo, Rafael José
Orientador(a): Mascarenhas, Nelson Delfino d'Ávila lattes
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de São Carlos
Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
Departamento: Não Informado pela instituição
País: BR
Palavras-chave em Português:
Área do conhecimento CNPq:
Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/488
Resumo: Computed tomography has been widely used for medical diagnosis, since it allows viewing of internal parts of a body without overlapping structures. However, recent studies indicate a certain risk of cancer to the patients who undergo this type of examination due the radiation dose to which they are exposed. The purpose of this work is to develop new _ltering techniques in the CT image space, in order to provide a better quality to the images acquired with low radiation exposure to the patient. For noise reduction, a new denoising technique is developed based on a pointwise Maximum a Posteriori (MAP). The noise is considered Gaussian with zero mean, as observed experimentally, and the variance is estimated considering two cases: signaldependent noise and signal-independent noise. For the a priori density of the signal, we used di_erent non-negative probability densities (re_ecting the fact the pixels of an image are non-negative). In another approach, the histogram of the images were segmented into unimodal parts and each segment was _ltered using the _lter based on the MAP criterion with the a priori density that best _ts it. After _ltering, the evaluation of the method is performed using the following criteria: Mean Square Error, Improvement in Signal to Noise Ratio, Universal Image Quality Index and Structural Similarity Index. The 2D _ltering results are compared with the results obtained by pointwise Wiener _lter.