Logo Recognition Using Shape Descriptors and Fusion of Multiple Classifiers

Abstract

          An important problem in the field of document image processing is the recognition of graphical items, such as company logos. Logos are mainly used by companies and organizations to identify themselves on documents. Different feature extraction and shape description methods in logo image recognition systems have been studied by several researchers. In this paper, two region-based shape descriptors: Zernike moment descriptors and Geometric moment descriptors are studied and compared. The performance of an ensemble of classifiers, each trained on different feature sets is also evaluated. The recognition results of the individual classifiers are compared with those obtained from fusing the classifiers' output, showing significant performance improvements of the proposed methodology.