A robust information hiding methodology in wavelet domain

With the proliferation of digital multimedia, information hiding techniques have become more attractive to the researchers. High bitrate information hiding is different from digital watermarking: the former one tries to hide relatively large amount of auxiliary information instead of just one or a few verification bits. This paper presents a novel Discrete Wavelet Transform (DWT) domain high bitrate information hiding algorithm. In the proposed algorithm, the coefficients within the approximation subband of the one-level Wavelet decomposition are grouped into vectors to embed information bits. Low-frequency coefficients have been chosen for information hiding due to their relatively large amplitudes and the corresponding smaller step size in JPEG2000 quantization. A mathematical model is proposed to predict the Bit-Error-Rate (BER) of the algorithm under JPEG2000 compression. The proposed algorithm has reached an initial channel capacity of 1/64 bits per pixel (bpp). With performance improvements, the channel capacity has been increased to 1/32bpp and 1/16bpp, at the price of slightly higher BER. Preliminary experiments indicate that the algorithm is robust and most of the hidden information survives JPEG2000 compression.