Research On Technique Of Digital Image Watermarking Based On Wavelet Transform

As the rapid development of computer and network technologies, peoplegradually stepped into the digital information age. The network become the main wayof distribution of literature works, so it attracted the attention of people to protect thecopyright of multimedia digital products, and many scholars began to study the digitalwatermarking technology which can embed the copyright mark into digital products.Based on the analysis of the basic theory of digital watermarking, this papermainly studies the wavelet domain digital watermarking algorithm. This paperproposes a DWT digital watermarking scheme based on quantization embedding. First,2layer wavelet decomposition is performed to the target image. Then, the secondarydecomposition coefficients-LH2and HL2are partitioned. Last, to embed thewatermark through quantifying gray mean of every image block. This method canchoose these image blocks whose coefficients are the first Mth largest, adaptively,where M is the size of the watermark. Simulation results show that this algorithm canensure a transparency, and resistant to noise, compression, and other conventionalsignal attacks. However, if the watermark image has been rotated or scaled, thewatermark detection will fail. At the same time, this method has blind checkingprobability.In order to make the algorithm can resist geometric attacks, the paper presents anidea that to carry out image correction through the feature points before watermarkdetection. Based on the comparative analysis of Harris operator and Harris-Laplaceoperator, this paper puts forward a modified Harris-Laplace operator, which canreduce the redundancy of extracted feature points. This new operator can quicklydetect robust feature points that has geometrical invariability, for that these pointscontains scale characteristics. Then these feature points can be used to correct image.Simulation results show that if the image correction is carried out withgeometric-attacked watermark images, the attacked image can back to the original oneapproximatively, so the watermark detection can success.