Comparison of wavelet based watermarking techniques Using SVD

Watermarking is a mechanism to provide security with audio or video. In this study we discuss video mechanism in which information can be hidden directly by modifying intensity values or frequency components of an image. Intensity values of an image can be modified directly, for modifying frequency components image need to transform image using different transformation techniques. Important properties of watermark include imperceptibility and robustness. Singular values of an image have very good stability when small values are added to the image, it does not effect the quality with great variation. Singular values correspond to intensity and singular vectors represent geometric characteristics. Here we made a study of hiding information using different hybrid watermarking techniques, SVD with DWT, RDWT,CWT at various levels of transformations in low, high, mid bands of frequencies. The quality of watermarked data and extracted watermark is calculated using PSNR and NCC by applying various attacks. Depending on the level of transformation and band in which watermark is embedded for some attacks the image was robust and for some it was not. Information can also be hidden in R G B or Y U V color spaces. Watermark extracted from three channels is studied, it was found that some of the channels were vulnerable to many attacks and some were not. Performance evaluation was made by applying various attacks.