Verification of Authenticity of Stamps in Documents

Classical ink stamps and seals used for authentication of a document content have become relatively easy to forge by the scan&print technique since the technology is available to general public. For environments where a huge volume of documents is processed, an automatic system for verification of authenticity of stamps has been developed in the scope of this work. The process of stamp authenticity verification naturally must be preceded by the phase of stamp detection and segmentation – a difficult task of Document Image Analysis. In this work, a novel method for detection and verification of stamps in color document images is proposed. It involves a full segmentation of the page, extraction of features and further classification of the segments by means of Support Vector Machines. The evaluation has shown that the algorithm is capable of differentiating stamps from other color objects in the document such as logos or text and also genuine stamps from copied ones.