Performance improvement of spread spectrum spatial-domain watermarking scheme through diversity and attack characterisation

We present a spatial-domain-watermarking scheme for data hiding and copyright protection of still images using the attack characterisation approach. A small number of bits are embedded onto an image in the spatial domain using a method similar to the direct sequence spread spectrum. Perceptual analysis is used to guarantee the imperceptibility of the watermark and proper scaling of the raw mask values maximises the robustness of the watermark, while keeping it imperceptible to the human senses. We add two types of watermarks: reference and robust. The reference watermark is used to assess the way, in which the marked signal has been modified, so that the robust watermark can be optimally extracted. The watermarking system requires two keys; one for generating the pseudo-noise sequence for spread spectrum modulation and another for embedding the two watermarks in a checkerboard arrangement, where the use of the key introduces spatial uncertainty about the location of the watermark bits. This approach has been implemented for a spatial diversity of 64 and 256. Simulation results reveal the effectiveness of the new approach.