Uso do método fuzzy c-means para segmentação de imagens dermatoscópicas de lesões de pele

• Autor(a) principal: Coelho, Talita Salles
• Data de Publicação: 2012
• Outros Autores: Fernandes, Marco Antônio Rodrigues [UNESP], Miot, Hélio Amante [UNESP], Yoriyaz, Hélio
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://www.rbfm.org.br/index.php/rbfm/article/view/195; http://hdl.handle.net/11449/140606
• topic: Dermoscopic images; Segmentation; C-means fuzzy; Imagens dermatoscópicas; Segmentação; Fuzzy c-means

This work describes a clustering method called c-means fuzzy, utilized for skin lesion dermatoscopic image segmentation. This type of segmentation is based on the method of region grow that differs from conventional clusterization methods by the use of the concept of fuzzy numbers, once that it is appropriate to deal with uncertainties referent to image regions (clusters). The method consists in calculating the pixel pertinence degree in relation to the regions that it can pertain, defined by a given neighborhood. In this work this method was applied in three images of skin lesion that are: two melanomas and one nevu; obtaining three clusters class to each image. These clusters were used to calculate two threshold values. These thresholds were applied in the binarization algorithm to image segmentation. With aim of verify the efficiency of the method, the segmented images through c-means fuzzy method was compared with same images segmented by Otsu algorithm. The segmentation obtained by the FCM algorithm was visibly better than that obtained by Otsu algorithm, this occurs due to the fuzzy numbers influence, where a pixel can pertain to more than one region, but with different pertinence degree.