Robust logo watermarking based on wavelet transform modulus maxima

Abstract

In this paper, a novel blind logo image watermarking technique is presented which can be used to protect the copyright of digital images. Our technique uses two different watermarks. One is a high capacity multi-bit watermark used to embed a logo and the other is a low capacity single-bit watermark used for the detection of geometrical attacks. Both watermarks are embedded into the red, green, and blue components of an RGB image, weighted based on human visual system considerations. Each non-overlapping watermark is embedded using a spread spectrum approach, based on a pseudorandom noise (PN) sequence and a unique secret key. Robustness against geometric attacks is achieved with the help of wavelet transform modulus maxima, which is shift invariant, and by embedding it in different sub-bands of the wavelet transform. The experimental results show that our proposed watermarking scheme is robust to geometric attacks such as rotation, scaling, and translation and compares favourably against existing techniques.