Processamento de fotomicrografias por meio da transformada wavelet starlet

• Ano de defesa: 2015
• Autor(a) principal: Siqueira, Alexandre Fioravante de [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Microscopia eletronica de varredura; Microscopia; Processamento de imagens; Visão por computador; Wavelets (Matematica); Scanning electron microscopes
• Link de acesso: http://hdl.handle.net/11449/126425; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/18-08-2015/000837544.pdf

Microscopies have been used for morphology evaluation of different materials structures. However, their results can be affected by several external factors. Image processing techniques can be used to attenuate these factors, improving the results. In this study we propose a method for segmentation of photomicrographs, denominated Multi-Level Startelet Segmentation (MLSS), based on the starlet wavelet transform. The choise of an optimal segmentation level is given by Multi-Level Optimal Segmentation (MLSOS), that uses MLSS results and the Matthews correlation coefficient (MCC). MCC compares the obtained segmentations and ground truth images, choosing the best segmentation for the input image. MLSS and MLSOS are evaluated using precision, recall and accuracy. Jansen-MIDAS, an open-source software from these methods, allows using MLSS and MLSOS by the end user. This software was employed in the separation of elements in images of different materials, namely gold nanoparticles in natural rubber samples and fission tracks in epidote crystals. In these applications, the proposed method presented accuracy higher than 87% for all test images