Using invariant image features for synchronization in spread spectrum image watermarking

A watermarking scheme is presented in which the characteristics of both spatial and frequency techniques are combined to achieve robustness against image processing and geometric transformations. The proposed approach consists of three basic steps: estimation of the just noticeable image distortion, watermark embedding by adaptive spreading of the watermark signal in the frequency domain, and extraction of relevant information relating to the spatial distribution of pixels in the original image. The just noticeable image distortion is used to insert a pseudo-random signal such that its amplitude is maintained below the distortion sensitivity of the pixel into which it is embedded. Embedding the watermark in the frequency domain guarantees robustness against compression and other common image processing transformations. In the spatial domain most salient image points are characterized using the set of Hilbert first-order differential invariants. This information is used to detect geometrical attacks in a frequency-domain watermarked image and to resynchronize the attacked image. The presented schema has been evaluated experimentally. The obtained results show that the technique is resilient to most common attacks including rotation, translation, and scaling.