Research On Binary Text Image Digital Watermarking

In the network information era, digitization of information and its wide spread on the internet make digital products gain an increasing application. In the meantime, the protection of copyright and the information security become a problem demanding prompt solution. A new technology, hypermedia digital watermark technology, by which we can embed messages into the digital products, was born to settle the problem. The messages we embed could be the information of author’s copyright, and thus can be a strong proof when faced with an illegal use of the products. A number of researches have been undertaken on the watermark embedded in colored images, videos and frequencies, in sharp contrast, little work was done on binary text images. However, as a matter of fact, texts are considered as the most important one among all kinds of media. Combined with some characteristics of the binary text image, makes the study of its digital watermark technology is relatively difficult.This article focuses on the algorithm of binary text image digital watermarking, which firstly introduced a series of related knowledge and background of digital watermarking technology for preparation, secondly summarized and reviewed the previous methods, then proposed an anti-print scan digital watermarking algorithm aiming at Chinese characters document. First step is text segmentation, and then according to the features of Chinese characters we pick up the proper pixels to flip, by doing so we embed our watermarking messages. In addition, we accomplished the extraction of our embedded messages in printed and scanned texts. The algorithm in this paper only needs a very small amount of flipping pixels but excellences at imperceptibility together with the ability of resisting print and scan. According to our experimental result, the algorithm applies to many typefaces of Chinese character, for instance, the accuracy of the algorithm when applying to the text which consists of small fourth word in italics 95.42% and 93.87%. for the text consists of five word in italics. After Reed-Solomon algorithm, the entire message can be obtained correctly.