Abstract
Traditional watermark scheme has encountered two serious problems. First, it is easy to attack watermark by methods which are specifically designed for its algorithm. Second, an image may be embedded many watermark simultaneously. We have no ideal which watermark was embedded first which one was embedded latter. Therefore, to determine truly owner of the content becomes impossible. Digital image copyright protection scheme is very different from traditional watermark. No modification of original image is needed. Only what we should do is extract feature of image and register to the third party with it. At the same time, we get a time stamp to prove at what time we register the content. When argument happened, the first registrant is truly legal owner of the image. The first registrant is defined who has the earliest time stamp. There are many ways proposed for feature extraction. But seldom feature remains unchanged under various attacks, especially geometric attacks. Our feature extraction method is called Mexican Hat Wavelet scale interaction. The extracted feature points can resist both geometric distortion and signal processing. Here come the problems about feature points pattern matching when we want to check an image registered or not. We observe that feature points have structural position shift. So our methods applied many inverse transformations to correct its distortion when suffered from attacks.