Melanoma is the deadliest form of skin cancer if left untreated. Incidence rates of melanoma have been increasing, especially among young adults, but survival rates are high if detected early. Unfortunately, the time and costs required for dermatologists to screen all patients for melanoma are prohibitively expensive. There is a need for an automated system to assess a patient's risk of melanoma using photographs of their skin lesions. Dermatologists could use the system to aid their diagnosis without the need for special or expensive equipment. One challenge in implementing such a system is locating the skin lesion in the digital image. Most existing skin lesion segmentation algorithms are designed for images taken using a special instrument called the dermatoscope. The presence of illumination variation in digital images such as shadows complicates the task of finding the lesion. The goal of this research is to develop a framework to automatically correct and segment the skin lesion from an input photograph. The first part of the research is to model illumination variation using a proposed multi-stage illumination modeling algorithm and then using that model to correct the original photograph. Second, a set of representative texture distributions are learned from the corrected photograph and a texture distinctiveness metric is calculated for each distribution. Finally, a texture-based segmentation algorithm classifies regions in the photograph as normal skin or lesion based on the occurrence of representative texture distributions. The resulting segmentation can be used as an input to separate feature extraction and melanoma classification algorithms. The proposed segmentation framework is tested by comparing lesion segmentation results and melanoma classification results to results using other state-of-the-art algorithms. The proposed framework has better segmentation accuracy compared to all other tested algorithms. The segmentation results produced by the tested algorithms are used to train an existing classification algorithm to identify lesions as melanoma or non-melanoma. Using the proposed framework produces the highest classification accuracy and is tied for the highest sensitivity and specificity.

Melanoma is the deadliest form of skin cancer if left untreated. Incidence rates of melanoma have been increasing, especially among young adults, but survival rates are high if detected early. Unfortunately, the time and costs required for dermatologists to screen all patients for melanoma are prohibitively expensive. There is a need for an automated system to assess a patient's risk of melanoma using photographs of their skin lesions. Dermatologists could use the system to aid their diagnosis without the need for special or expensive equipment. One challenge in implementing such a system is locating the skin lesion in the digital image. Most existing skin lesion segmentation algorithms are designed for images taken using a special instrument called the dermatoscope. The presence of illumination variation in digital images such as shadows complicates the task of finding the lesion. The goal of this research is to develop a framework to automatically correct and segment the skin lesion from an input photograph. The first part of the research is to model illumination variation using a proposed multi-stage illumination modeling algorithm and then using that model to correct the original photograph. Second, a set of representative texture distributions are learned from the corrected photograph and a texture distinctiveness metric is calculated for each distribution. Finally, a texture-based segmentation algorithm classifies regions in the photograph as normal skin or lesion based on the occurrence of representative texture distributions. The resulting segmentation can be used as an input to separate feature extraction and melanoma classification algorithms. The proposed segmentation framework is tested by comparing lesion segmentation results and melanoma classification results to results using other state-of-the-art algorithms. The proposed framework has better segmentation accuracy compared to all other tested algorithms. The segmentation results produced by the tested algorithms are used to train an existing classification algorithm to identify lesions as melanoma or non-melanoma. Using the proposed framework produces the highest classification accuracy and is tied for the highest sensitivity and specificity.

Orientador: Nilton Nobuhiro Imai / Resumo: O uso do UAV no sensoriamento remoto é uma área crescente de conhecimento e pressionou o desenvolvimento de câmeras multi e hiperespectrais leves, que poderiam ser usadas embarcadas em UAV. Como as informações espectrais dependem das condições de iluminação da cena, as imagens adquiridas por UAV exigem pesquisa para avaliar o processamento de imagens para melhor adaptar os conceitos já estabelecidos às imagens orbitais. Portanto, a correção radiométrica é de fundamental importância para a extração de dados de imagens com alta confiança, uma vez que se sabe que o fator de refletância é uma função da estrutura geométrica, ângulo solar e propriedades ópticas. Nesse sentido, foram desenvolvidas metodologias para corrigir imagens das diferenças de iluminação, utilizando as Funções de Distribuição de Refletância Bidirecional e os Modelos de Correção Topográfica, também conhecidos como correção de iluminação. Este trabalho utiliza a câmera hiperespectral Rikola a bordo de um UAV em latitudes tropicais. Ele avalia como a anisotropia pode influenciar a variabilidade na reflectância dos alvos e como os modelos de correção topográfica podem ser aplicados, usando a micro topografia, para atenuar esses efeitos. Três testes foram realizados para estudar i) as geometrias de visada da câmera hiperespectral Rikola e a disponibilidade de dados fora do Nadir, ii) a variação do fator de anisotropia entre os alvos nas geometrias de visada e iii) modelos de correção de microtopografia para corrigi... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: Use of UAV in remote sensing is a growing area of knowledge and pressed the development of lightweight multi and hyperspectral cameras, which could be used on board of UAV. Since the spectral information depends on the lighting condition of the scene, UAV acquired images demands research to assess image processing to better adapt the concepts already stablished to orbital images. Therefore, the radiometric correction is of main importance for data extraction from imagery with high confidence, once it is known that the reflectance factor is a function of geometric structure, solar angle and optical properties. In this regard, methodologies were developed to correct images from illumination differences, using the Bidirectional Reflectance Distribution Functions and the Topographic Correction Models, also known as illumination correction. This Work uses Rikola hyperspectral camera onboard of a UAV in tropical latitudes. It assesses how the anisotropy can influence variability in reflectance of targets, and how topographic correction models can be applied, using micro topography, to attenuate these effects. Three tests were done to study i) the view geometries of Rikola Hyperspectral camera and the no-Nadir data availability, ii) the variation of Anisotropy Factor between targets in the many view geometries and iii) Microtopography correction models to correct illumination differences using a highly detailed DSM (10 cm and 3 cm) to assess the micro relief. We applied the correcti... (Complete abstract click electronic access below) / Mestre