Imagens de radar de abertura sintética (SAR) são tipicamente corrompidas pelo ruído \"speckle\" que também degrada imagens geradas por ultra-som, laser, etc. Esta tese propõe algoritmos de filtragem baseados na abordagem \"maximum a posteriori\" (MAP) para redução de \"speckle\" em imagens SAR. Na derivação dos filtros MAP, para imagens obtidas por detecção linear, são utilizadas as distribuições (condicionais) Rayleigh e raiz quadrada de gama na regra de Bayes como modelos para o ruído \"speckle\" em imagens SAR obtidas em amplitudes com 1 e múltiplas visadas, respectivamente, e usadas várias distribuições para o modelo \"a priori\". Toda a formulação dos algoritmos tem por base o modelo multiplicativo que constitui o modelo mais adequado ao \"speckle\". Propõe-se ainda neste trabalho a combinação dos filtros MAP formulados com o algoritmo k-médias e com a técnica de crescimento de regiões, como forma de melhoria da abordagem de filtragem proposta. Os resultados de filtragem foram avaliados segundo critérios (medidas) de melhoria da relação sinal-ruído e perda de resolução. O primeiro critério avalia a redução da intensidade do ruído \"speckle\" sobre regiões homogêneas e para avaliar a perda de resolução decorrente da filtragem é proposta uma nova técnica baseada na transformada de Hough. Os algoritmos foram testados em imagens artificialmente contaminadas por ruído \"speckle\" e em imagens SAR reais apresentando estatísticas Rayleigh e raiz de gama. Os resultados obtidos mostram a melhoria que proporcionam os algoritmos de filtragem MAP, especialmente quando combinados com o classificador k-médias e com a técnica de crescimento de região. O uso da técnica de crescimento de região reforça a conclusão de que o uso de vizinhança estatisticamente mais semelhante ao pixel ruidoso melhora a estimação dos parâmetros de filtragem. As medidas de desempenho e validação dos algoritmos MAP permitiram concluir que os filtros com distribuições \"a priori\" Gaussiana, gama, chi-quadrado e beta apresentaram melhores resultados de filtragem em relação aos demais modelos \"a priori\" quando comparados ao filtro de Kuan e com a técnica de \"wavelets\" para a classe de imagens utilizadas / Synthetic aperture radar (SAR) images are typically corrupted by speckle noise, which also degrade images produced by laser beams, ultrasound, etc. This thesis proposes filtering algorithms based on the \"maximum a posteriori\" (MAP) approach, to reduce speckle in SAR images. To derive the MAP filters for linearly detected images we assumed the multiplicative model for the speckle and used the conditional density functions in the Bayes rule following a Rayleigh and square root of gamma for one-look and N-looks images, respectively, and several different \"a priori\" densities. The MAP filters are combined with the k-means classifier and region growing tools to improve the proposed filtering approach. Measures evaluating both the signal-to-noise improvement and resolution loss due to filtering are computed. To assess the improvement brought by the proposed algorithms we evaluate them with respect to signal to noise ratio and edge preservation. The former is a classical way to evaluate the speckle strenght reduction over homogeneous areas and the latter is a new proposed technique based on the Hough transform that measures distortions at the edges produced by the speckle MAP filtering algorithms. The qualitative analysis of the MAP proposed algorithms includes the methods based on the curvature and wavelets . The algorithms were applied to simulated noisy speckled images and real SAR images with statistics of linearly detected images with one-look and N-looks. The obtained results demonstrated the improvement brought by the speckle MAP filtering algorithms, specially when combined with the k-means clustering algorithm and with the region growing approach. This region growing approach reinforces the conclusion that the use of a neighborhood whose pixels have statistics similar to the noisy pixel provides a better estimation for filtering. The evaluating measures point out that the MAP filters whose \"a priori\" models are the Gaussian, gamma, chi-square and beta presented better results than the other \"a priori\" models proposed in this thesis, the Kuan filter and the wavelets filter, for the class of images that were tested

Tese de doutoramento. Engenharia Mecânica. Faculdade de Engenharia. Universidade do Porto. 2007

Thesis (M.S.)--Oregon Graduate Center, 1976.

Zugl.: Karlsruhe, Univ., Diss., 2009 / Hergestellt on demand. - Zusätzliches Online-Angebot unter http://uvka.ubka.uni-karlsruhe.de/shop/isbn/978-3-86644-418-8

Title from PDF of title page (University of Missouri--Columbia, viewed on Feb 15, 2010). The entire thesis text is included in the research.pdf file; the official abstract appears in the short.pdf file; a non-technical public abstract appears in the public.pdf file. Dissertation advisor: Dr. Ping Yu. Vita. Includes bibliographical references.

It is known that a phase-perturbed object field from a fractal-like rough surface illuminated by a partially coherent beam generates speckle on the image plane. In this works, aberration effects on Gaussian speckle on the image plane in both perfectly and partially coherent systems are theoretically and experimentally investigated. Theory shows that the second order statistics of Gaussian laser speckle are independent of odd-functional aberrations, but they do affect Gaussian speckle contrast in a partially coherent system. Furthermore, it is theoretically derived that field statistics of Gaussian laser speckle generally become non-circular Gaussian due to aberrations, and an aberration effect is asymptotically ignorable for very weak or strong roughness. A brute force simulation method is introduced for non-Gaussian speckle in a partially coherent imaging system, where speckle irradiance is calculated from a quasi-monochromatic extended incoherent source. The source is modeled as a collection of independent point sources distributed on a regular grid. The partially coherent speckle pattern is calculated from the incoherent sum of coherent speckle patterns in the image plane generated from each point source. Speckle contrasts from a brute force model show good agreement with theoretical and experimental results. It is determined that non-Gaussian speckle contrast is strongly dependent on Hurst exponent of fractal rough surfaces using brute force simulations. The concept of a contributing object area at a fixed image point effectively explains the speckle contrast dependency. Measuring spherical aberrations using Gaussian laser speckle is discussed as one of applications and future works of the present study.

Power transformers play an important role in electric networks. Copper windings wrapped in kraft paper and immersed in a tank of mineral oil are their main composition. Oil-impregnated paper is a very common insulator in power transforms, due to its durability and endurance. Kraft paper, a wood-based fiber from cellulose, can gradually degrade due to the ambient effects such as high-voltage electric field, temperature, and moisture in the normal working condition of transformers. Aged paper is more prone to experience electric breakdown which finally results in transformer failures. It is essential to monitor paper insulation condition. Conventional methods for evaluating kraft paper are mainly based on chemical analyses which require sampling of the surrounding oil and are usually time-consuming and expensive. Investigation on and classification of aged kraft paper with four different levels of thermal ageing using a potentially fast, simple, and inexpensive optical setup based on optical speckle is the main subject of this thesis. Speckle interference patterns from a laser diode source have been captured by a CCD camera. Classification based on textural features shows very good discrimination between ageing classes making this technique promising for industrial applications.

Thesis (Ph. D.)--Oregon Graduate Center, 1988.

Thesis (M.S.)--Rochester Institute of Technology, 1990. / Typescript. Includes bibliographical references (leaves 135-143).

Thesis (Ph. D.)--University of Washington, 1996. / Vita. Includes bibliographical references (p. [125]-135).