Research On The Text Digital Watermark Algorithm

In recent years, the developments of information technologies bring us so many conveniences. However, they also introduce a great deal of trouble at the same time. People can copy and tamper digital products more easily without permission of their copyright owners. In this situation, as an effective approach which is applied to protect security of multimedia product, digital watermarking has already become skilled day by day. Currently, watermarking research is mostly focused on the image and audio, while fewer are focused on the text. As the most widely used information storing and convening media, the application of the text is further and wider than that of the image, video, audio. Digital watermarking should be widely applied in the field of copyright protection and content authentication of the text.In this paper, we first introduced some related knowledge about digital watermarking and text watermarking. For the texts, Brassil et. al. proposed the marking scheme of line shifting and centroid detection method, which were both based on space domain. In Reference 12, Huang Hua et. al. simplified and improved this algorithm. They proposed a new marking scheme of line shifting and the corresponding centroid detection method. The new scheme can provide double marking spaces than the old one does, because the new scheme gave up the use of control lines. The computation is greatly reduced. Since it only needs to check the relative distances of each adjacent line's centroid, the new scheme can be blindly detected. The main work of this paper is as follows: (1).A property of the improved method in Reference 12 has been proved. Under the same influence of Gaussian noise, the influence of centroid of text lines is the same as that of the original method.(2).The probability of false detection of the improved method has been calculated. Compared with the original method, the probability of false detection of improved method has increased, which was also pointed out in Reference 12. However, the probability of false detection is very small, so the improve method can be applied in practice.(3).In addition, because of the probability of large error for short text line, Reference 30 proposed the codirection expansion strategy. Thus, we can apply it to the detection method of Reference 12. By the expansion of the initial text line, we constructed a regenerative simulation text line centroid sequence, which can effectively reduce the probability of false detection. The probability of false detection of the new method is smallest among the original Brassil centroid detection method, the improved one and our proposed.