Image watermarking in wavelet domain is a challenging problem as it includes a proper selection of wavelet transforms as well as sub-bands for accurate embedding and recovery of watermarks.However, selection of a suitable wavelet transform for multiple image watermarking is an area of great interest. Therefore, in this work, a multiple watermarking technique using combination of redundant discrete wavelet transforms (RDWT), non-subsampled contourlet transform (NSCT), set partitioning in hierarchical tree (SPIHT), and singular value decomposition (SVD) is proposed. The proposed scheme is aimed to achieve high capacity, robustness, and imperceptibility. The security aspect of the proposed method is enhanced by using Arnold transform. Since RDWT and NSCT are both shift invariant in nature, therefore they are suitable for multiple image watermarking. Furthermore, the proposed method provides flexibility in selection of appropriate sub-bands for watermark embedding and recovery. Experimental results and analysis reveals that the proposed technique gives maximum PSNR, NC, and SSIM value up to 40.97 dB, 1, and 0.9994, respectively. Furthermore, the performance estimation of our technique is found superior to previously reported techniques under consideration.
KEYWORDSArnold transform, non-sub sampled contourlet transform, RDWT, SPIHT, SVD
INTRODUCTIONRecently, secure transmission of digital data over open/public networks is one of potential challenges. 1 Therefore, there is a great requirement to secure the data transmission over networks. Digital image watermarking is one of promising method for protecting digital data and is widely used for copyright protection and secure communication. 2-4 Image watermarking provides secure transmission of digital data by embedding it into the cover image for content authentications, copyright protection, 5 etc.Discrete wavelet transform (DWT) has been widely used for image watermarking. 6 However, the major drawbacks associated with the DWT are its shift sensitivity and poor directionality that can be avoided by using redundant wavelet transforms. 7 The existing literatures over image watermarking clearly states that the redundant wavelet transforms are suitable for single and multiple images watermarking 7-10 as these transforms do not use down sampling operation in implementation. Thus, no data/information loss occurs during forward and inverse transformation. Furthermore, redundant discrete wavelet transform (RDWT) 11 and non-subsampled contourlet transform (NSCT) 12 are shift invariant. This property enables us to have better reconstruction of images and found useful in many image processing applications. Furthermore, NSCT provides rich directional information, which can be used in embedding single as well as multiple watermarks after forward transformation. 12 It is evident that the performance of combined transform domain techniques (two or more than two) is better than individual one. 7 Thus, being motivated by above facts, an image watermarking technique in NSCT domain using...