Anais do X SIBGRAPI (1997) #AND15


Pesquisa de Imagens em Banco de Dados por Semelhança de Cor

Everest Mathias, CAA-UFF
Aura Conci, CAA-UFF

Abstract. In image retrieval based on their color content, the distance between color histograms may be defined as a match measure. However, this measure operates on high dimensional space and is computionally expensive, function of the number of color. Moreover retrieval by color similarity, given a used-sampled image, is particularly challenging, owing to the difficulty to derive a measure that conforms the human perception. The degree of matching achieved on image querying depends on, at least: (1) the number of histogram bins used, (2) the underlying color space and (3) the metric used for histogram matching. This paper describes three different approaches for solving the querying by color similarity problem. Examples from a prototype system illustrate the effectiveness of each approach.

[ full version | contact authors | demo ]