Segmentação de texturas em imagens digitais utilizando wavelets redundantes

• Ano de defesa: 2015
• Autor(a) principal: Dobler, Joyce Aline Duarte [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Computação - Matematica; Wavelets (Matematica); Imagens de sensoriamento remoto; Cores; Computer science - Mathematics
• Link de acesso: http://hdl.handle.net/11449/138485; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/07-04-2016/000860785.pdf

The texture analysis plays an important role in many applications, such as analysis of medical imaging, industrial inspection, remote sensing, etc. The main applications that use the analysis of textures is the segmentation. In particular, the study related to the textures segmentation spans for more than five decades, however, it continues to be topic of enormous interest. Because of the variety of contexts and situations in which the term texture is used, there isn't formal definition of the concept. Intuitively, texture is related to tonal variations (gray level) in images, and can be that considered local discontinuities that are generally associated with edges. The edge detection is generally seen as a process that reduces the amount of data that represents a sign, however, the wavelet transform is a powerful tool that uses scale, directionality and the local energy distribution of the edges, and provides a complete and stable representation of a signal, without loss of information. In this study, we studied and adapted the texture segmentation method developed by Pagamisse using the dyadic wavelet transform with four directions, including starlet wavelet transform in order to improve their performance. The method agregates and combines the responses of the redundant wavelets transform (invariant by translation) in scales and orientations, providing the best segmentation of textures. The efficiency of the method constructed was proven by applying to mosaics of textures and comparing the results obtained with the results of other texture segmentation methods found in the literature