EXTRACTION OF MULTIPLE SHAPES IN COLOR TRADEMARK IMAGES

摘要

With the ever increasing number of registered trademarks, the task of trademark office is becoming increasingly challenging to ensure the uniqueness of each trademark logo. The current computer-aided search system based on manual assignment of codes employed by many trademark offices has several limitations, which leads to considerable attention in employing content-based image analysis in trademark image search and registration. Among the many features associated with content-based analysis, shape plays the utmost important role in describing trademark logos. It is thus very important to have all important shapes in trademark logos identified and extracted for analysis. In this paper, we present a novel method for the extraction of multiple important shapes in trademark logos based on color segmentation and shape analysis. Experimental observations on several hundreds trademark images of various kinds shows promising results.