PPGINF PÓS-GRADUAÇÃO EM ENGENHARIA DA INFORMAÇÃO FUNDAÇÃO UNIVERSIDADE FEDERAL DO ABC Phone: Not available http://propg.ufabc.edu.br/ppginfo

Banca de DEFESA: MIGUEL BOZER DA SILVA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : MIGUEL BOZER DA SILVA
DATA : 26/05/2021
HORA: 09:00
LOCAL: Remoto
TÍTULO:

Deep Convolutional Networks For Super-Resolution Image Reconstruction


PÁGINAS: 101
GRANDE ÁREA: Engenharias
ÁREA: Engenharia Elétrica
RESUMO:

The need to obtain high resolution images exists in several areas, such asmedicine, astronomy, security and monitoring. However, digital cameras usually donot have the desired high resolution or some external factor can reduce the qualityof an image that is is being studied. Thus, several studies have emerged in the areaof Super Resolution of images, in order to obtain high resolution images from lowresolution images using softwares for this task. The most used methods today arebased on only a single input image single image super resolution (SISR) and, usingthe deep learning techniques, it is possible to train a network with objective ofobtaining high resolution images with characteristics similar to the training set.The method proposed for this work uses two branches of convolution thatdiffer in the way that the image has its scaling factor increased for the first timein the proposed architecture. In one of the branches, the transposed convolution isused and in the other, the bicubic interpolation. After that, in each of the branches,the feature maps pass through𝑛up and down-projection units and their output areconcatenated. One last convolution layer is applied to obtain the high resolutionimage. With this type of structure, we can see that the scenario with two branchesobtain better performance in terms of PSNR and SSIM when compared to thesame scenario with only one branch when we use𝑛= 1and𝑛= 4units of up anddown-projection units. For the scenario with𝑛= 6there were no differences betweenthe models of one and two branches.


MEMBROS DA BANCA:
Presidente - Interno ao Programa - 1761107 - RICARDO SUYAMA
Membro Titular - Examinador(a) Interno ao Programa - 2334927 - ANDRE KAZUO TAKAHATA
Membro Titular - Examinador(a) Externo à Instituição - MAGNO TEOFILO MADEIRA DA SILVA - USP
Membro Suplente - Examinador(a) Interno ao Programa - 2090028 - FILIPE IEDA FAZANARO
Membro Suplente - Examinador(a) Externo à Instituição - LEVY BOCCATO - UNICAMP
Notícia cadastrada em: 30/04/2021 07:54
SIGAA | UFABC - Núcleo de Tecnologia da Informação - ||||| | Copyright © 2006-2021 - UFRN - sigaa-2.sigaa-2