Studies On The Robust Watermarking Technologies For The Digital Image

As an important part of multimedia information,the digital image has a wide application in the modern era.With the popularity of smartphones and the development of mobile Internet,the number of digital images has exploded and the digital images are almost everywhere in our lives.As one kind of digital property,digital images have economic value.But with the development of the technology,it is becoming easier to copy or edit images.As the technology to embed information into images,digital image watermarking is an effective approach to protect the copyright or integrity of the images.In the current environment,there are many different kinds of attacks for the digital images.The traditional digital image watermarking schemes maybe invalid.Therefore,designing the new robust digital image watermarking schemes has the special practical sense.In this paper,the author mainly studies the robust watermarking for digital image,and designs a variety of watermarking schemes for the common attacks at the current stage.This study covers spatial domain watermarking methods,transform domain watermarking methods and compressed domain watermarking methods.It still involves blind watermarking methods and non-blind watermarking methods.It also mentions one reversible lossless watermarking method and a few lossy watermarking methods.There are four main innovations:(1)In order to cope with the common attacks such as image collage,crop,rotation and scaling,a compound watermarking scheme is proposed.Under these attacks,the size of the original image not only changes,but also part of the image is tampered.Therefore,common watermarking schemes may fail under such attacks.In order to solve this problem,the author designed a watermarking scheme which contains the noise template watermarking and a IWT-DCT based watermarking.The noise template watermarking method has good robustness and imperceptibility,and the scaling rate and rotate angle can be calculated by this method,too.But the embedded information cannot be freely edited when we using the noise template watermarking method.For the watermarking method based on IWT-DCT,the embedded information can be freely edited,and this method can also resist crop and translation.But this method cannot resist scaling and rotate attack.Thus these two methods can be complementary for embedding,and the robustness is strongly improved.Through the experiment results,the watermarking information can be detected and extracted after the image is attacked by the ways mentioned above.(2)A reversible lossless watermarking scheme based on histogram and visual cryptography is designed for the Block Truncation Coding images.In this scheme,one improved method of histogram shift is used.This method has good imperceptibility and this is a lossless embedding method.After the extraction,the image can be turned to the original state.This watermarking scheme is designed to resist tampering attack and make the watermark information more secure.The scheme has a high information embedding capacity,and at the same time the visual cryptography improves the security of the watermark.The visual cryptography can divide the watermark information into different transparent shares,and only have enough shares,the watermark information can be recovered.Thus the security of the whole scheme is enhanced.The designed watermarking scheme based on BTC compression domain has high security,and it is robust to tampering attacks under the BTC data format.(3)The watermarking schemes based on histogram and low frequency feature construction are proposed for solving the problems which are image secondary compression and scaling in mobile applications.The traditional watermarking schemes maybe invalid under these attacks.The original watermarking method based on histogram has good imperceptibility and robustness,but this method has low capacity.Thus we propose an improved method based on the histogram which has larger embedding capacity.In this scheme,author also proposed a low-frequency feature construction method to construct low-frequency features in the spatial domain.Author designed one way to calculate the weights about pixels in the different position,and the pixel which has the biggest weight is the most suitable to modify.After the modification,the low frequency feature of the image is enhanced,and the image will be smoother.The low frequency feature construction method can improve the robustness and optimize the imperceptibility.In this section,the author proposed a non-blind watermarking scheme and a blind watermarking scheme to cope with different practical application scenarios.The non-blind watermarking scheme has good robustness,but the robustness of the blind watermarking scheme is weaker.(4)A watermarking scheme for vector quantization compressed domain images is proposed in this section for extracting valid information under attacks such as micro-angle rotation,scaling,white noise and other attacks.When using the codebook which has worse quality,the robustness of the watermarking will decrease.Thus,in this section,the author focus on the way of generating codebook,and the quality of the codebook is enhanced and the computational time is shorter.The author proposed a fast high-quality codebook design method based on particle swarm optimization(PSO).And the method also contains the pooling,reverse cast and K-means process.The pooling process can reduce the number of the training vectors for PSO,and PSO can get the coarse cluster results.The reverse cast process can cast the coarse cluster results to the original vector training set.By using K-means for fine-turning,the good cluster centers will be generated,and the final codebook has a good quality.The codebook which has a higher quality will lead to a more robust watermarking scheme.And fast codebook generating process has broad applications.