Blind image watermarking via exploitation of inter-block prediction and visibility threshold in DCT domain

摘要: Inter-block relation in DCT domain is exploited to attain blind image watermarking.A neural network offers an appropriate prediction for relative comparison.The between a coefficient and its leads binary embedding.Embedding strength adjusted subject just-noticeable difference model.The proposed scheme exhibits superior robustness imperceptibility. In this paper, the backward-propagation (BPNN) technique (JND) model are incorporated into block-wise discrete cosine transform (DCT)-based achieve effective watermarking. To form block structure domain, we partition host non-overlapped blocks of size 8×8 then apply each separately. By referring certain coefficients over 3×3 grid blocks, BPNN can offer adequate predictions designated inside central block. The watermarking turns out be process adjusting relationship intended their JND. Experimental results show that able withstand variety processing attacks. Compared with two other schemes also utilize inter-block correlations, one apparently imperceptibility under same payload capacity.