Abstract: Seal imprints have been extensively used in many oriental countries. The automatic identification of seal imprints is a very difficult problem of pattern recognition. Every phase of the automatic identification of seal imprints is studied carefully and a corresponding series of algorithms is presented. In order to make a distinction between surplus parts of strokes in forgery seal imprint and surplus parts of seal paste in genuine seal imprint, and make a distinction between missing parts of strokes in forgery seal imprint and missing parts of seal paste in genuine seal imprint, several difference images and image processing methods are proposed, i. e., full difference image, inner difference image, surplus difference image, missing difference image, etc. On such a basis, an automatic seal imprint identification scheme is presented. Experimental results confirm that the proposed scheme are feasible for practical applications.