@MastersThesis{Monego:2021:ReImAs,
               author = "Monego, Vinicius Schmidt",
                title = "Restaura{\c{c}}{\~a}o de imagens astron{\^o}micas utilizando 
                         redes neurais, wavelets e regulariza{\c{c}}{\~a}o",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "2021",
              address = "S{\~a}o Jos{\'e} dos Campos",
                month = "2021-03-04",
             keywords = "problemas inversos, restaura{\c{c}}{\~a}o de imagens, 
                         ru{\'{\i}}do, regulariza{\c{c}}{\~a}o, transformada wavelet, 
                         inverse problems, image restoration, noise, regularization, 
                         wavelet transform.",
             abstract = "Neste trabalho s{\~a}o estudadas diferentes t{\'e}cnicas de 
                         restaura{\c{c}}{\~a}o de imagens, envolvendo m{\'e}todos de 
                         regulariza{\c{c}}{\~a}o, filtros wavelets e redes neurais. Mais 
                         especificamente, as t{\'e}cnicas escolhidas foram 
                         regulariza{\c{c}}{\~a}o por Tikhonov, regulariza{\c{c}}{\~a}o 
                         por entropia, regulariza{\c{c}}{\~a}o da varia{\c{c}}{\~a}o 
                         total, filtro de Wiener, filtragem wavelet, redes neurais 
                         convolucionais e filtro neural multiescala. As imagens de 
                         interesse s{\~a}o imagens astron{\^o}micas, provenientes da 
                         fonte HubbleSite, que disponibiliza imagens do telesc{\'o}pio 
                         espacial Hubble sob uma licen{\c{c}}a compat{\'{\i}}vel a 
                         dom{\'{\i}}nio p{\'u}blico. As imagens s{\~a}o degradadas com 
                         ru{\'{\i}}do gaussiano de desvio padr{\~a}o de 5%, 15% e 25%. A 
                         performance de cada um dos m{\'e}todos de restaura{\c{c}}{\~a}o 
                         {\'e} avaliada atrav{\'e}s das m{\'e}tricas: NRMSE (Normalized 
                         Root-Mean-Square Error Erro M{\'e}dio Quadr{\'a}tico 
                         Normalizado), PSNR (Peak Signal-to-Noise Ratio Raz{\~a}o de pico 
                         sinal-ru{\'{\i}}do) e SSIM (Structural Similarity Index Measure 
                         Medida do {\'{\i}}ndice de similaridade estrutural). ABSTRACT: 
                         In this work, different image restoration techniques are studied, 
                         involving regularization methods, wavelet filters, and neural 
                         networks. More specifically, the techniques chosen were 
                         regularization by Tikhonov, regularization by entropy, total 
                         variation regularization, Wiener filter, wavelet filtering, 
                         convolutional neural networks and multiscale neural filter. The 
                         images of interest are astronomical images from the HubbleSite 
                         source, which makes images from the Hubble space telescope 
                         available under a license compatible with the public domain. The 
                         images are degraded with standard deviation Gaussian noise of 5%, 
                         10% and 15%. The performance of each of the restoration method is 
                         evaluated using the following metrics: Normalized Root- 
                         Mean-Squared Error (NRMSE), Peak Signal-to-Noise Ratio (PSNR) and 
                         Structural Similarity Index Measure (SSIM).",
            committee = "Stephany, Stephan and Campos Velho, Haroldo Fraga de (orientador) 
                         and Kozakevicius, Alice de Jesus (orientadora) and Queiroz, 
                         Gilberto Ribeiro de and Silva Neto, Ant{\^o}nio Jos{\'e} da and 
                         Shiguemori, Ana Paula Abrantes Castro",
         englishtitle = "Astronomical image restoration using neural networks, wavelets and 
                         regularization",
             language = "pt",
                pages = "101",
                  ibi = "8JMKD3MGP3W34R/44ARSBE",
                  url = "http://urlib.net/ibi/8JMKD3MGP3W34R/44ARSBE",
           targetfile = "publicacao.pdf",
        urlaccessdate = "2024, Apr. 28"
}