Estudo do método SPH na resolução de problemas de inpainting
Carregando...
Data
2020-09-29
Autores
Ferreira, Leticia Braga Berlandi [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O inpainting digital, ou retoque digital, é o processo de reconstrução das partes faltantes ou danificadas em uma imagem, tendo como objetivo principal a recomposição gradual dessas áreas, recuperando assim o seu aspecto natural da maneira mais nítida possível. No mundo digital, o inpainting refere-se à aplicação de algoritmos sofisticados que reconstituem, por meio de alguma estratégia específica de preenchimento como síntese de texturas, uso de equações diferenciais parciais ou modelos de representação esparsa, as partes perdidas ou corrompidas da imagem. Assim, neste trabalho é proposto um estudo do método Smoothed Particle Hydrodynamics para retoque digital de imagens, o que inclui comparações com outros métodos correlatos da literatura. Os fundamentos desse método encontram-se inseridos na teoria de interpolação e em sistemas de partículas, tendo já sido reformulado com sucesso para tratar outras aplicações na frente de Processamento de Imagens como remoção de ruídos e animações gráficas, cuja literatura serviu de apoio para a realização deste trabalho. Além disso, esse método não faz uso de malhas para estimar derivadas, o que o torna bastante interessante e versátil no tratamento do problema de inpainting, pois a solução numérica do problema é construída sem que a conectividade entre as partículas dessa aproximação seja preestabelecida. Dessa forma, tem-se como meta obter a harmonização de uma imagem, utilizando a perspectiva humana como referencial para avaliar se o resultado obtido é satisfatório ou não.
Digital inpainting (or digital retouching) is the process of reconstructing the missing or damaged parts within an image, having as main objective the gradual recomposition of these areas, therefore recovering the natural aspect of the image as clearly as possible. Computationally speaking, the term inpainting refers to the application of sophisticated algorithms that reconstruct, by applying some specific filling strategy such as texture synthesis, partial differential equations or sparse representation models, the missing data or corrupted portions of an image. Bearing this in mind, we propose to apply the so-called Smoothed Particle Hydrodynamics (SPH) method for digital image inpainting purposes. The background of such approach lies in interpolation theory and particle systems, which have successfully been reformulated to treat other applications in the field of Image Processing, such as noise removal and graphic animations. Moreover, SPH does not make use of intricated meshes to estimate derivative values, which makes it very interesting and versatile in the treatment of the inpainting problem, since the numerical solution of the related problem is built without imposing any connectivity scheme among the particles. Therefore, the goal of this master degree research is to achieve the visual harmonization of the inpainted image, by using the human vision as a benchmark to assess whether or not the recovered image is satisfactory.
Digital inpainting (or digital retouching) is the process of reconstructing the missing or damaged parts within an image, having as main objective the gradual recomposition of these areas, therefore recovering the natural aspect of the image as clearly as possible. Computationally speaking, the term inpainting refers to the application of sophisticated algorithms that reconstruct, by applying some specific filling strategy such as texture synthesis, partial differential equations or sparse representation models, the missing data or corrupted portions of an image. Bearing this in mind, we propose to apply the so-called Smoothed Particle Hydrodynamics (SPH) method for digital image inpainting purposes. The background of such approach lies in interpolation theory and particle systems, which have successfully been reformulated to treat other applications in the field of Image Processing, such as noise removal and graphic animations. Moreover, SPH does not make use of intricated meshes to estimate derivative values, which makes it very interesting and versatile in the treatment of the inpainting problem, since the numerical solution of the related problem is built without imposing any connectivity scheme among the particles. Therefore, the goal of this master degree research is to achieve the visual harmonization of the inpainted image, by using the human vision as a benchmark to assess whether or not the recovered image is satisfactory.
Descrição
Palavras-chave
Retoque digital de imagens, Smoothed particle hydrodynamics, Interpolação, Sistemas de partículas, Image inpainting, Interpolation, Particle systems