Document and text watermarking detection algorithm based on image processing

Digital watermarking is widely believed to be a valid means to protect the copyright of documents. Watermark detection plays an important role during the watermarking, because, during the distribution of the document, the printing-scanning-and-copying attacks will make the watermarked words expanded, cankered and spotted which make the detection even more difficult. It is necessary to develop the robust and anti-attacking watermarking detection algorithms. A document watermarking detection algorithm based on image processing is introduced for getting anti-attacking and indiscernible method of watermark. In Chinese characters, there are many strokes that are curly, and the degree of the curl is not easy to be discernible. So it is more transparent of embedding watermark in this way. The printing-scanning-and-copying attacks will change the thickness and length of strokes largely and change the degree of curl little. So detecting the degree of curl could have the potential ability of anti-attacking. Experimental results show that the algorithm is robust to some attacks, and the watermarking can be detected accurately. The correct rate of extraction of strokes is 99.96%, and the correct rate of detection is 86.1%.