Development of Improved Attack Resistant Image Watermarking Technique

Abstract

Digital watermarking is a method that prevents unauthorized distribution of copyrighted material by providing means of proving ownership of the material. This thesis introduces a watermarking technique based on lifting wavelet transform (LWT), schur and singular value decomposition (SVD). Generalized arnold transform (GAT) scrambling is applied on a grayscale watermark logo which works for improving the security of the watermark. The colored image is decomposed into high level and low level sub-bands using LWT, first schur decomposition is applied on watermark as well as the selected High-High (HH) sub-band to get the upper triangular matrices and then watermark is embedded in the image using SVD. “Peak-Signal-to-Noise-Ratio” (PSNR) and “Normalized Correlation” (NC) are the two metrics through which we can measure the perceptual quality and similarity of the extracted watermarks. The proposed scheme is compared under various attacks including noising attacks (salt & pepper, multiplicative etc.), geometric attacks (crop, translation, projection etc.), image processing attacks (histogram equalization, gaussian smoothing, contrast enhancement etc) were also applied on the watermarked images to check the performance of the scheme under attacks. The results show that the proposed scheme performs better in terms of quality and robustness in various attacks than the existing scheme.