O presente trabalho tem como finalidade a selecionar de um dado conjunto de atributos, o subconjunto que melhor classifique pixels de imagens frontais de faces humanas. Inicialmente foi utilizado o método de Análise de Componentes Principais (PCA, do inglês Principal Component Analysis) para selecionar os atributos mais importantes para a tarefa. O conjunto de atributos, chamados aqui de descritores, são de 4 tipos: descritores de primeira ordem, de segunda ordem, Momentos Invariantes e Multiescala. Todos obtidos de imagens frontais de faces humanas convertidas para escala de cinza. Por fim, utilizamos redes neurais de função de base radial (RBF, do inglês Radial Basis Function) para, utilizando o subconjunto escolhido, segmentar as imagens em duas, três e quatro classes, onde temos boca, nariz, olhos pele, pelos e plano de fundo.. Palavras-chave: Segmentação de Faces. Reconhecimento de Faces. Seleção de Atributos. Descritores de Textura. Classificação de Imagens

We live in a society, where, each day, more and more information is produced. Much of this information comes as images by various means such as security cameras, MRI machines and cameras for ordinary users. Thus, we have entered a time where 36 billion photos are added each year in one of the largest in the world and social networking sites and places with restricted access has granted access through of the image displayed by advanced devices which make the recognition of the individual.Techniques for face recognition has attracted the attention of many researchers. In recent years, techniques for face recognition have attracted the attention of many researchers.Several algorithms have been developed, from computer vision applications in recognition systems using small cameras, to automatic access control systems. Segmentation is one process that can help in the recognition of human faces. the segmentation of human faces tries to extract correctly from an image important components of the face, such as eyes, nose, mouth, hair, face and others. However, several variations on the images, such as lightning conditions and facial expressions tend to difficult such work. In this work, we intend to analyze and propose the best subset(s) of features with respect to the image segmentation of faces in frontal position. Keywords: Face Segmentation. Face Recognition. Feature Selection. Textural Features. Image Classification