Performance loss of dirty-paper codes in additive white gaussian noise and jitter channels

Abstract

Dirty-paper codes have been used extensively as the basis for new code design for digital image watermarking due to their interference rejection propeny. Random Bending Attacks still pose a great danger to watermark survival and the theoretical basis for the assessment of their impact on decoder performance is yet to be laid. Meanwhile, recent research results point out to the use of a new channel model named Additive White Gaussian Noise and Jitter (AWGN&J) channel as a means to assess the impact of imperfect synchronization in the performance of the watermark decoder. In this paper, we show that the capacity of the Ideal Costa Scheme is greatly reduced in the presence of imperfect synchronization and that its host signal interference rejection propeny does not hold anymore in the presence of such attack. Moreover, we show that arbitrarily increasing watermark power does not help to mitigate the effect of imperfect synchronization.

Application of QIM Based Audio Watermarking for Synchronizing Detection in the Air Environment

“…In [17], the watermark was repeatedly embedded at the edge of the audio signal by removing and adding a portion of the audio signal while preserving the pitch and viewing the pitch-invariant TSM as a special form of random cropping. In [18], the invariance of the binomial wavelet transform with linear scaling was used to design audio watermarking by modulating the wave shape.…”

On the Capacity Equivalence with Side Information at Transmitter and Receiver

“…Following the pioneering work of Costa, several extensions of DPC have been studied, e.g., colored Gaussian noise [5], arbitrary distributions of S [6] and deterministic sequences [7]. In [8], DPC has been applied to the AWGN and jitter channel and the rate loss due to imperfect synchronization at the decoder has been evaluated. DPC has also been extensively applied to watermarking or information embedding [9], [10] applications and for the Gaussian broadcast channel [11], [12].…”

A scaled signal plus noise model for digital watermarking. Application to time jitter

“…In this section, after a short presentation of the AWGN&J channel, AWGN&J desynchronization effects are investigated differently: (i) using common Inter Symbol Interference (ISI) assumption commonly known in the communcation theory, and (ii) using the model (6). Both models are finally compared.…”

“…an AWGN&J channel is an AWGN channel in which the signal is, in addition to the gaussian noise, randomly sampled. More precisely, the receiver has to decide on th epresence of a watermark based on attacks using the AWGN&J model already exist, [5][6][7]. However, in these works, the desynchronization noise is almost usually considered uncorrelated with the watermarked signal.…”