A Robust Digital Image Watermarking in Hybrid Frequency Domain

Abstract

Image watermarking is a method to hide the secret information in a host image for copyright protection of watermark data during the transmission by means of insecure channel. The proposed scheme protects our data with adaptive level of visual quality and robustness against signal processing and geometric attacks. The proposed method divides the host image into four non-overlapping segments labelled as sub-images, DWT is applied on each sub images and then block based DCT is applied on mid frequency channels LH and HL of discrete wavelet transform. Embedded matrix is formed using hybrid transformed coefficients where matrix elements are chosen from the localized two mid frequency coefficients of each block in DCT. SV Decomposition is applied on embedded matrix to factorize it into singular values, left and right singular vectors and embed the scrambled watermark image along with scaling factor in singular value matrix. This repetition of watermark data in each sub-image reduces the PSNR values of the watermarked image. Despite this proposed scheme scales down PSNR value, changing the scaling factor favours to adjust the PSNR to the acceptable level and withstand the signal processing attacks such as JPEG compression and geometrical attack such as rotation, translation. Compared to the other method, the proposed scheme gives better correlation coefficient value for above mentioned kinds of attacks and also provide adaptive PSNR for imperceptibility on watermarked image.