Face Segmentation Based On Texture Classification And Heuristic

Título: Face Segmentation Based On Texture Classification And Heuristic

Autores: Laboreiro, V. R. S.; Chrisóstomo, H. B.; Araujo, T. P. de; Maia, J. Everardo Bessa

Resumo: Facial image segmentation is treated as a texture classification problem. We use supervised learning to classify image pixels into C classes, each class corresponding to a face part to be segmented. Four types of features are used: first order gray level parameters (2 features), textural measures (4 features), multi-scale features (1 feature) and moment invariant features (2 features), totaling nine features. A Radial Basis Function Neural Network (RBFNN) is used to classify image pixels into 11 regions (or classes): right and left eyes, right and left eyebrows, nose, mouth, right and left ears, face, hair, and background. Satisfactory qualitative and quantitative results were obtained and presented.

Palavras-chave: Face segmentation; sum and difference histogram; radial basis function neural network; heuristic

Páginas: 5

Código DOI: 10.21528/CBIC2011-15.2

Artigo em pdf: st_15.2.pdf

Arquivo BibTex: st_15.2.bib