Mapa auto-organizÃvel com campo receptivo adaptativo local para segmentaÃÃo de imagens

AUTOR(ES)

Diogo Cavalcanti Costa

DATA DE PUBLICAÇÃO

2007

RESUMO

A new self-organizing map with variable topology is introduced for image segmentation. The proposed network, called Local Adaptive Receptive Field Self-organizing Map (LARFSOM) is based on the Self-organizing Map (SOM) and Grow When Required (GWR). The model main features are: adaptive final number of nodes, variable topology, new node insertion based on similarity threshold calculated by a local receptive field, non-significant node removal after training, fast convergence and low computational cost for training stage. LARFSOM is capable of both color and border segment images. The color segmentation stage is performed by the clustering process during the LARFSOM training stage, followed by an image color quantization procedure. For border segmentation RBF (Radial Basis Function) nodes are included to determine the border pixels using previously learned information of LARFSOM, this two-stage model is called LARFSOM-RBF. Moreover, the proposed model is capable of saving the trained network and the spatial pixel image information in a variant format of indexed BMP. A ZIP compression is them applied to result in a compact file. LARFSOM is compared with SOM, FS-SOM and GNG, self-organizing maps used for color segmentation of real world images of increasing complexities. Image processing techniques and the JPEG format are also used for comparisons. LARFSOM reached higher color palette variance and better 3D RGB color space distribution of learned data for the trained images them the other models. LARFSOM was tested to segment images with different degrees of complexity showing promising results

ASSUNTO(S)

redes neurais image processing color segmentation image compression representaÃÃo/compactaÃÃo de imagens grow when required mapas auto-organizÃveis (soms) funÃÃo de base radial (rbf) segmentaÃÃo por cor neural networks processamento de imagens crescer quando requerido (gwr) segmentaÃÃo de imagens ciencia da computacao self-organizing map border segmentation image segmentation radial basis function segmentaÃÃo por borda

Documentos Relacionados

• SegmentaÃÃo de imagens geradas por laser scanning aerotransportado para delimitaÃÃo de Ãrvores individuais em Ãreas de reflorestamento de eucaliptos
• Sistema para segmentaÃÃo de imagens por agrupamento hierÃrquico baseado no comportamento superparamagnÃtico do modelo de Potts
• SegmentaÃÃo e extraÃÃo das caracterÃsticas de defeitos em imagens radiogrÃficas de juntas soldadas e tubulaÃÃes
• SegmentaÃÃo de sentenÃas manuscritas atravÃs de redes neurais artificiais
• SegmentaÃÃo e classificaÃÃo de padrÃes visuais baseadas em campos receptivos e inibitÃrios