Indexing technical symbols using ridgelets transform

Abstract

This paper presents a new use of the Ridgelets Transform as a multiscale method for indexing linear symbols in graphic documents. The Ridgelets Transform is useful to detect linear singularities in images. Therefore, it can be used to get a good representation of linear symbols. Moreover, as it belongs to the Wavelet family, this representation can be useful at several scales of detail. Thus, the proposed scheme targets to detect linear singularities at different scales, using the Ridgelets properties to group similar symbols at rough scales. In this way, we can use Ridgelets representation when searching in large symbol databases by finding similar symbols at rough scales and discriminating them at finer scales. (c) Springe r-Verlag [2004.