An Analysis of Wavelet Based Dual Digital Image Watermarking Using SVD

Abstract

Intellectual properties are being obtained and reproduced easily. This raises a serious question on the ownership of images. Hence content owners and service providers seek protection of digital multimedia through watermarking. Though watermark is inserted in host image it becomes difficult to prove ownership due to various attacks affecting the image. Hence watermarking should be as robust as possible maintaining the fidelity. As wavelet based image watermarking resembles the human visual system, it is gaining vital importance in protecting copyright information. In this paper a hybrid scheme based on DWT and SVD is implemented. Two watermarks are inserted in host image one by one. After decomposing the primary watermark image into four bands, SVD is applied to each band, and the same secondary watermark is embedded by modifying the singular values. In this way new watermark is generated which is then embedded into host image. Analysis of the method is done after performing different attacks such as rotation, addition of Gaussian and Poisson noise, average filtering etc. on the watermarked host image. The paper also implements dual image watermarking schemes based on DWT and SVD separately. Correlation coefficient of extracted secondary watermark is 0.23481 in SVD based technique whereas in DWT based technique it is 0.74443 for Gaussian noise attack. On the other hand for DWT-SVD based method it is 0.88701. Thus performance of combined DWT-SVD dual image watermarking technique is proven optimised in comparison with the previously mentioned two methods.