On improving the iterative watermark embedding technique for JPEG-to-JPEG watermarking

We previously proposed an iterative technique called iterative watermark embedding (IWE) for watermark embedding in JPEG compressed domains. The proposed technique embeds the watermark successfully when the quality factor of the JPEG image is high, but when the quality factor is low, not all the bits of the watermark can be embedded. This is due to uneven distribution of capacities in different regions of the image. Moreover, the computation complexity of the iterative loop is high. In this paper we improve the technique using shuffling to equalize the capacities of different regions. Also, the iterative loop is replaced by a direct approach to reduce the computational complexity. We call this direct approach Direct JPEG Watermarking Embedding (DJWE). A Human Visual System (HVS) model is used to prioritize the DCT coefficients to achieve good visual quality. Experimental results show that the watermarks are robust against Gaussian noise attack and JPEG transcoding attacks.