Studies Of Template Based Blind Image Watermarking

With the growth of electronic products and network, the application of digitalimage is more and more popular. The convenience of storing, sharing digital images isboth a merit and a demerit. On one hand ,it provide a fast way of transferring images,on the other hand , it put forward challenge for copyright protection.Digital watermarking is the right technique to full fill the need of protecting copy-right for digital image. It embed some information(text, icon, even a sign notifyingthe watermark exists) without remarkably affect the visual effect of the image. Theembedded information can be extracted after the digital image is intentionally or unin-tentionally modified. In practical situation, the original copy of digital image may notbe available in extracting embedded information. That is called a blind extraction.In this article, blind extraction for large digital image is studied. The principalprocessing methods a large digital image may encounter are cropping and scaling, wehope to find a robust blind image watermarking scheme that is robust to scaling andcropping. Traditional watermarking schemes simulates cropping by replacing a por-tion of the image with blank, and scaling by scale back after scaling to some smallerscale. These ways of simulation keeps the synchronization method(size, relative posi-tion, rotation angle etc.) and does not meet the practical situation.After about 2000, watermark schemes that aims at the synchronization problembecome a hot spot in the field of blind watermarking. 3 major ways become the mainstream. First is Log-Polar mapping based approach, aims at transforming the transla-tion, scaling, rotation into invariant or simple translation in some transform domain.The second is salient feature based approach, aims at embedding and extracting water-mark on the basis of finding salient feature points which are robust against cropping,scaling and rotation in image. The third is template based method, aims at embeddingsome anchor points or blocks into image that can be used to restore the image from the transformation performed on it.In this article, we firstly improve a traditional space-domain watermarking scheme,making it more robust to JPEG compression and scaling. Based on this scheme andexisting schemes that aims at synchronization problem, we proposed two schemes.One is Harris-Laplace feature point based blind watermarking schemes, the other istemplate blind watermarking scheme based on Lab color space. Both schemes canprovide robustness against scaling and cropping with synchronization information re-moved. Besides, we made detailed analysis on the performance of these two schemes,and try to summarize the merits and demerits of salient feature based approach andtemplate based approach.