Digital Image Watermarking Algorithm Based On Wavelet Transform

With the development of the Internet and multimedia technologies, the copyright protection of digital contents has become one of the serious problems. The Digital watermark technologies are new research directions for protecting copyrights of digital contents such as image and video. The Digital watermark technologies include spatial domain watermark and frequency domain watermark. Because of the merit of being more tolerant to attacks and image processing, frequency domain watermark has become research focus currently. Because discrete wavelet transform (DWT) can well match the JPEG2000 standard and has the characteristic of multi-resolution analysis which accords with human visual system (HVS), the wavelet-based watermarking has been developed into the primary method of frequency domain watermark..Generally digital watermarking technology has two elementary characteristics. One is Unobtrusiveness, the other is robustness. The former means that watermark embedded will not make quality of digital information worse to be found by attackers. The latter means that digital watermarking should be supposed to resist attacks against common signal processing. Embedding position and Embedding intension factor are crucial to robustness and unobtrusiveness.In this paper, we make an analysis to the current wavelet-based watermarking methods, find that the input image will be separated into one low frequency part (LL) and some high frequency parts (HL,LH,HH) by wavelet multi-resolution analysis. The low frequency part reflects the rough information and the majority of image energy concentrate here. The high frequency parts denote detail information about diverse scales and resolutions. The current watermarking methods in frequency domain can be divided into two classes. One is embedding watermark in low frequency part, the other is embedding watermark in high frequency parts. But both classes of methods have deficiencies. Because mankind eyes are more sensitive to low frequency parts than to high frequency parts, The first class of methods may make the quality of digital image worse and decrease the Unobtrusiveness of the watermark.. The better Unobtrusiveness of watermark can be obtained by using the second class of methods, however, the watermark embedded is easy to be destroyed by some signal processes such as losing compression.In order to get over the disadvantages of the two classes of methods above, we propose a wavelet-based watermarking method which embeds watermark in middle frequency parts. By this method, a well trade-off between Unobtrusiveness and robustness has been achieved by adjusting the Embedding intension factor according to the characteristics of HVS. The results of experiment show that the method is effective to both gray digital images and color digital images.