Trademark shape recognition using closed contours

In this paper, a new trademark recognition method using closed contours is proposed. The developed method can be used to retrieve similar trademarks, even those with irregular shapes, so as to simplify and speed up the manual trademark examination process. Trademarks are first decomposed into complete sets of elementary closed contours, each of which is then encoded as an angle-code string according to chain-code information. A two-step string-matching algorithm can then be employed to compute similarities between closed contours. Finally, the maximum and average terms of the similarities between contours are integrated into the whole trademark similarity measure. After that, the similarity measure can be the criterion for retrieving similar trademarks. More specifically, since the proposed method to encode closed contours is insensitive to translation, rotation and size variations, our method has the ability to detect infringement cases in which trademarks are changed in translation, rotation, or size. Moreover, our method can cope with the special resemblances such as inclusion, mirror and partial resemblances. Experimental results are presented to show the feasibility and practicability of the proposed approach. (C) 1997 Elsevier Science B.V.