Research On Typical Histogram-based Digital Watermarking Algorithm

Thehistogram of an image can be well used in watermarking algorithms robust to geometric distortions because the histogram is only related to the pixel number of each grayscale, not the position of pixels. Meanwhile, digital products may also be attacked by some traditional operations in the distribution procedure, such as compression.Most of existing histogram-based watermarking methods focused on the geometric invariance feature more while paying less attention to the robustness to some common attacks. Giving consideration to both attack categories, following schemes are proposed in this paper:1. For the problems found in the classic histogram-based method, a new histogram modification scheme is proposed. Meanwhile, we combine block dividing and histogram statistics for the first time in watermarking scheme. We use the mean pixel value of divided blocks to calculate histogram and the mean square error to select blocks to be modified based on HVS. We can obtain a good trade-off between robustness to geometric attacks and common image processing attacks by adjusting the block size.2. Based on the multi-frame feature of videos, we group the frames and use block dividing and histogram modification on each block. Besides, parity-check and staggered statistics are used in the scheme. The scheme proposed can improve the extraction accuracy rate effectively even test videos are under geometric and transcoding attacks.3. Considering the robustness problems found in spatial histogram-based method, another watermarking algorithm based on the frequency domain is proposed. Combining the property of DCT and the QIM method, we modify the coefficient of full-frame DCT to embed bits. Besides, group marks are added in embedding bits for majority vote in case of frame dropping in practical circumstances.