Toward a Better Understanding of DCT Coefficients in Watermarking

DCT is widely used in watermarking, but the effect of different coefficients on robustness and fidelity remains unclear. For example, there is only one DC coefficient, but many low frequency coefficients, so error correcting coding can be applied to improve the robustness of the low frequency coefficients, and in this situation, how about the robustness between a single DC coefficient and all the low frequency coefficients? In this paper, the characters of different DCT coefficients are studied when error correcting coding is used. Experimental results and analyses show that even error correcting coding technique is used among all the lower frequency coefficients when watermark is embedded, the robustness of all the lower frequency coefficients is still inferior to the single DC coefficient. Besides, under the same peak signal to noise ratio, the visual effect of different coefficients differ from each other, and the DC coefficient gets the worst visual effect.