In this paper, we propose an image watermarking system that is highly robust against various attacks without perceivable image degradation. The cover image is first discrete wavelet transformed (DWT), and then the low and middle subbands are divided into wavelet blocks. A selective watermark embedding method is used in which a DWT block is chosen for watermark embedding only when its coefficients clearly indicate the block polarity. Instead of the original image, a key is used in the watermark extraction to indicate the locations where watermark bits are embedded. The key is generated by a Tri-state Exclusive OR (TXOR) operation on the randomized watermark and the randomized DWT coefficients of the original image. Finally, a deadzone evacuation procedure is performed to ensure an adequate noise margin. If a DWT coefficient is very close to the polarity threshold, e.g., the median, then it will be forced to shift to the positive or the negative end of the deadzone depending on its polarity. Simulation results show that the key method proposed herein achieves excellent performance for Checkmark non-geometric attacks, such as filtering, compression, and copy attacks. The proposed scheme is also robust for image cropping at different positions.
|頁（從 - 到）||368-377|
|期刊||Proceedings of SPIE - The International Society for Optical Engineering|
|出版狀態||已出版 - 2002|
|事件||Security and Watermarking of Multimedia Contents IV - San Jose, CA, United States|
持續時間: 21 1月 2002 → 24 1月 2002