%0 Conference Proceedings
%4 sid.inpe.br/banon/2002/11.29.11.39
%2 sid.inpe.br/banon/2002/11.29.11.39.40
%@doi 10.1109/SIBGRAPI.2001.963042
%T Multi-bands image analysis using local fractal dimension
%D 2001
%A Conci, Aura,
%A Nunes, Eldman de Oliveira,
%E Borges, Leandro Díbio,
%E Wu, Shin-Ting,
%B Brazilian Symposium on Computer Graphics and Image Processing, 14 (SIBGRAPI)
%C Florianópolis, SC, Brazil
%8 15-18 Oct. 2001
%I IEEE Computer Society
%J Los Alamitos
%P 91-98
%S Proceedings
%1 SBC - Brazilian Computer Society
%K local fractal dimension,texture analysis,multi-bands images,multispectral bands.
%X An important application of fractals is image texture analysis. The main aspect of fractal geometry used in this application is the concept of fractal dimensions to characterize the texture-scaling behavior. A new idea is presented here: the use of fractals of texture identification in multiband image analysis. This is not a simple extension of the usual characterization of multifractal from its local dimension in gray-level images. It is related to examine the interrelationship among the image representation in bands. Moreover, each band can be seen as a set in the 3D space, wich means that its fractal dimension may present results between 2 and 3. Consequently if two bands are considered in one gathering, their structure is a set in the 4D space and its fractal dimension may present results between 2 and 4. For multi-bands image the fractal dimension upper bound can be even larger. In this work, two propositions to handle the multi-band combination are presented. Both defined for whatever combination of bands. As far as we know, no similar approach has been considered until now.
%@language en
%3 91-98.pdf
%O The conference was held in Florianópolis, SC, Brazil, from October 15 to 18.