Geometric distortion correction based robust watermarking scheme in LWT-SVD domain with digital watermark extraction using SVM

Abstract

With the advent of technology, digital image watermarking turns to be an effective technique to protect digital images from illegal usages and manipulation. In digital image watermarking, one of the major challenges is to provide robustness against geometrical attack maintaining adequate level of imperceptibility and security. In this work, a robust digital image watermarking scheme is proposed based on the combination of lifting wavelet transform(LWT) and singular value decomposition(SVD). To achieve better correlation between the extracted and original watermark, SVM-based binary classification approach is integrated in watermark extraction. In the proposed technique, geometric distortion correction based approach is incorporated with SVM based binary watermark detection to achieve improved robustness against de-synchronization attack. The 3-level LWT is performed on the cover image where horizontal (HL) sub-band is chosen for binary watermark insertion. The training and testing patterns are formed using an optimized set of features along with the singular values of corresponding blocks. In case of de-synchronization attacks, geometrical distortion correction is required before performing watermark extraction. In the detection process, the geometric distortion parameters of the attacked watermarked image are estimated by the geometric correction method. This algorithm provides high robustness against both the geometrical and non-geometrical attacks. It has been observed that the algorithm gives average imperceptibility of ~42.27 dB with the watermark capacity of 512 bits. Experimental results suggest that the proposed watermarking scheme provides significant improvement in terms of robustness and security of the watermark. The subjective analysis also suggests that proposed scheme provides improved performance over some of the recent existing techniques.