Estudo de técnicas de registro e reconstrução de imagens aplicadas ao problema de super-resolução

Detalhes bibliográficos
Ano de defesa: 2015
Autor(a) principal: Arthur Faria Porto
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Minas Gerais
UFMG
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://hdl.handle.net/1843/BUBD-A7NF7Y
Resumo: This work presents an experimental evaluation of dierent image registration and reconstruction techniques applied to the Super-Resolution problem. The Super-Resolution problem aims to generate a high resolution image from a set of low-resolution images of the same scene. The Super-Resolution approach can be divided into two phases: registration and reconstruction. After registering the low resolution images on a single high resolution grid, the reconstructionphase combines these images into a single image of high resolution. Four image registration methods (Keren, Marcel, Lucchese and Vandewalle) and ve image reconstruction methods were evaluated (IBP, TV, RSR, NC and G-PMSR). The objective is to assess, among the considered methods, which is the best combination of techniques able to generate a high resolution image with a better expression of the observed scene. Initially experiments were conducted in a controlled environment, generating articially low resolution images, to evaluate the performance of registration and reconstruction techniques considered. After the selection of the excelled methods in the initial experiment, a new experiment, with real low resolution images was performed to validate the combination of chosen methods. The results demonstrate thefeasibility of using the combination of the Keren and G-PMSR methods in Super-Resolution problems. The high resolution images generated were able to retrieve details of the scene that were not perceptible with the low resolution images alone.