Research On Image Fingerprinting Technology Of Copyright Detection

Image fingerprinting is a kind of copyright protection technology and it has been used in the digital image works. As a newly emerging technology, image fingerprinting has been paid more and more attention. At present, the proposing of a safe and efficient fingerprinting algorithm has become the research focus. The main work of this thesis is summarized as follows:Firstly, image features extraction is a key technology for Image fingerprinting. A novel image feature extraction scheme and a new way for image representation proposed in this paper, which can improve the ability to discriminate different images while also recognize same images with affine attacking which including rotation, scale and translation. Several functions are combined together to process the preprocessed original images, which will produce feature vectors of the original images. Experimental results on ORL face database prove that the feature extracted by this scheme can discriminate different images while also recognize the same images with affine attacking. Therefore, this algorithm is feasible in practical applications of copyright protection, image Retrieval, face recognition and et al.Secondly, it takes Windows XP SP3as the development platform, Visual C++2008and Oracle10g database as development tools of experimental system to design and build the image feature extraction system based on invariant feature of trace transform. The interface of the user program is friendly and the operation of program is easy. Meanwhile, the system has good expandability. The paper selected the ORL database and another100image as image library for more effective test. In order to test the algorithm’s affine attacks resisting capability, this paper used the StirMark benchmark for attacking images. The results show that the scheme can well withstand geometrical attacks. This scheme is feasible for extracting image features.This paper completes the research on image features extraction based on invariant feature of trace transform and adds reasonable image measuring methods of comparability, thus it can well withstand affine attacks including rotation, scale and translation smoothly.