Abstract: A geometric attack resisting and watermarking blind extraction implemented algorithm is proposed based on compressive sampling(CS) techniques and the SIFT feature.The first copyright watermarking is spectrum spread and embedded in DWT domain of NCST’s lower frequency band.The second watermarking is authentication watermarking.It is generated by compressive sampling of the first watermarking,and stored in IPR database as zero-watermarking.In the process of watermarking extraction,firstly,SIFT feature template is acquired through zero-watermarking,and it veri-fies the integrity of watermarked image,locates and restores alterations.Then the image is calibrated with scale features and coordinate relationship of SIFT feature points so as to update watermarking extracting locations.The simulation shows that,the proposed algorithm has huge watermarking capacity and favorable transparency.It is robust to ordinary attacks as well as several geometric attacks.

Key words: Digital watermarking,Compressive sampling(CS),Zero-watermarking,Scale invariant,Geometric calibration