A fast watermarking algorithm with enhanced security using compressive sensing and principle components and its performance analysis against a set of standard attacks

摘要

An algorithm for watermarking of digital images is proposed in this paper which utilizes Compressive Sensing (CS) with Principle Components (PCs) to achieve robustness, speed and security. CS is applied on PCs of watermark image to get the CS measurements. The singular values of these CS measurements are embedded with a scale factor into the HL subband of the cover image. The generated watermarked image contains three-layer security: one from PCs and other two from CS measurements. To recover PCs from CS measurements, a convex optimization tool, namely, the Orthogonal Matching Pursuit (OMP) is employed. Experiments are performed on both types of cover images; one with more low frequency components and another with more high frequency components. The algorithm offers state-ofthe art values of robustness and security in presence of different checkmark attacks like geometrical, non-geometrical and JPEG compression. A comparison of robustness of proposed algorithm with existing algorithms reveals that the proposed algorithm outperforms for most of the noise attacks. The performance of proposed algorithm with different wavelet families (e.g., orthogonal, biorthogonal, symmetric and asymmetric) are compared in terms of robustness and execution time. Such comparison may be helpful in selecting a suitable wavelet for a class of cover images in presence of checkmark attacks. The Haar wavelet performs better for geometric noise attack whereas Bior6.8 and Sym8 for non-geometric and JPEG compression type of noise attacks. The execution time of proposed algorithm with Haar wavelet is found to be minimum for all checkmark attacks. Moreover, it is quite less as compared to Optimization based methods and close to the other watermarking technique used for H.264 video standard.