A minimum distortion data hiding technique for compressed images

We present a blind data hiding method for JPEG compressed images which minimizes the perceptual distortion due to data embedding. The proposed system presents a number of op. tions to the encoder to cast the given hidden bits in the compressed content signal. The perceptual distortion cost of each option is calculated from the parameters available to the encoder such as the original image, quantization error due to compression and the Just Noticeable Distortion (JND) levels of the original image derived through an empirical human visual system model. The encoder selects the option with the minimum Jh’D cost to cast the hidden hits. By the definition of blind decoding, the decoder should be able to extract the hidden bits without any side information on the option selected or the parameters available to the encoder. The decoder of the proposed system uses simple binary addition on the received transform coefficients to extract the hidden bits blindly. System performance is examined by computer experiments at different compression levels and at different embedding bitrates.