TheWeighted Gradient: A Color Image Gradient Applied to Morphological Segmentation

Franklin Cesar FloresAirton Marco PolidorioRoberto de Alencar Lotufo

This paper proposes a method for color gradient computationapplied to morphological segmentation of colorimages. The weighted gradient (with weights estimatedautomatically), proposed in this paper, applied in conjunctionwith the watershed from markers technique, providesexcelent segmentation results, according to a subjectivevisual criterion. The weighted gradient is computedby linear combination of the gradients from eachband of an image under the IHS color space model. Theweights to each gradient are estimated by a systematicmethod that computes the similarity between the image tocompute the gradient and an 'ideal image', whose histogramhas an uniform distribution. Several experimentswere done in order to compare the segmentation resultsprovided by the weighted gradient to the results providedby other color space metrics, also according to a subjectivecriterion, and such comparison is present in thispaper.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: