Genetic Algorithm for Weights Assignment in Dissimilarity Function for Trademark Retrieval

Abstract

Trademark image retrieval is becoming an important application for logo registry, verification, and design. There are two major problems about the current approaches to trademark image retrieval based on shape features. First, researchers often focus on using a single feature, e.g., Fourier descriptors, invariant moments or Zernike moments, without combining them for possible better results. Second, even if they combine the shape features, the weighting factors assigned to the various shape features are often determined with an ad hoc procedure. Hence, we propose to group different shape features together and suggest a technique to determine a suitable weighting factors for different shape features in trademark image retrieval.

In this paper, we use a supervised learning method for finding the weighting factors in the dissimilarity function by integrating five shape features using a genetic algorithm (GA). We tested the learned dissimilarity function using a database of 1360 monochromatic trademarks and the results are promising. The retrieved images by our system agreed well with that obtained by human subjects and the searching time for each query was less then 1 second.