Vector image retrieval based on approximation of Bezier curves with line segments

This paper presents a similarity retrieval system for vector images such as trademarks. The proposed system retrieves vector images from a database using a sketch given by user as a query. The sketch is converted into a vector image format as a pre-processing step for the similarity retrieval. The similarity evaluation is proceeded by the following three steps: (1) approximating Bezier curves representing contours into line-segments, (2) extracting feature points from the line-segments, and (3) measuring the similarity between the sketch and each vector image in the database by computing the difference between the spatial distributions of feature points with the EMD (Earth Mover´s Distance). In order to evaluate the retrieval performance of the proposed system, we have conducted experiments using 32 sketches for querying a database containing 1141 trademarks. The retrieval results have shown that to 23 queries in the 32 queries, the correct trademarks are 1st-ranked.