Research On Robust Image And Video Watermarking Algorithms Based On Scale Invariant Feature Transform

The rapid development of the information industries helps people share the multimedia information,including images and videos,conveniently by the Internet,and also provides the convenience for the infringement,such as piracy,illegal copy,and transmission.Thus,copyright protection issues of digital images and videos have become the urgent problem to be solved.Digital robust watermarking techniques emerge at the right moment.Aiming at problems in the robust watermarking,including the selection of watermark embedding location,the balance between imperceptibility and robustness of the watermark,as well as the poor robustness to geometric attacks,this thesis studies the robust image and video watermarking algorithms based on scale invariant feature transform(SIFT).The main work done is as follows:(1)To solve the problem with the selection of watermark embedding location and obtain the better watermarking imperceptibility,a robust image watermarking algorithm based on SIFT feature points embedding,all phase biorthogonal transform(APBT),and singular value decomposition(SVD)is designed,where SIFT is applied to select the feature points.The watermark information is embedded sparsely into the singular values of the coefficients matrices obtained by APBT.Experimental results show that,in the premise of better imperceptibility,the proposed algorithm has robustness to the common attacks and combination of attacks.(2)To resist the geometric attacks in different degrees,an anti-geometric attacks robust image watermarking based on SIFT is designed.By extracting the feature points information of watermarked images and attacked images by SIFT,the feature matching is realized to correct the failure of watermark synchronization.The watermark information is embedded into the singular values in three-level DWT sub-bands.Experimental results show that the algorithm has preferable robustness to common attacks,geometric attacks,and combination of attacks,which solves the problem of poor robustness to geometric attacks in conventional robust watermarking.(3)To improve the watermarking robustness to geometric attacks,a novel affine SIFT(ASIFT)based robust image watermarking,which is resistance to geometric attacks,is proposed.ASIFT is applied for the image geometric attacks correction.Compared with the method based on SIFT,experimental results show that the proposed scheme has better robustness to scaling and translation attacks.(4)To improve the video watermarking robustness to rotation attack,an anti-rotation robust video watermarking based on SIFT is proposed.Aiming at the rotation attack,a strategy of attack correction based on SIFT is applied.Moreover,by scene change detection,motion frames are selected for watermark embedding,and watermarks are embedded into the two-level sub-bands.Experimental results show that this method is robust to noise attacks and rotation attacks in different degrees.