Fixed-Distortion Orthogonal Dirty Paper Coding for Perceptual Still Image Watermarking

A new informed image watermarking technique is proposed incorporating perceptual factors into dirty paper coding. Due to the equi-energetic nature of the adopted codewords and to the use of a correlation-based decoder, invariance to constant value-metric scaling (gain attack) is automatically achieved. By exploiting the simple structure of orthogonal and Gold codes, an optimal informed embedding technique is developed, permitting to maximize the watermark robustness while keeping the embedding distortion constant. The maximum admissible distortion level is computed on a block by block basis, by using Watson’s model of the Human Visual System (HVS). The performance of the watermarking algorithm are improved by concatenating dirty paper coding with a turbo coding (decoding) step. The validity of the assumptions underlying the theoretical analysis is evaluated by means of numerical simulations. Experimental results confirm the effectiveness of the proposed approach.