Histogram-Based Difference Expansion for Reversible Data Hiding with Content Statistics

Reversible data hiding can be classified into two major research branches, one is by use of the histogram of original image, named the H scheme, and the other is by manipulating pixel relations between adjacent pixels, called the difference expansion (DE) scheme. Both schemes have their own advantages and drawbacks, but the inherent characteristics of original image are seldom considered in existing works. We integrate the advantages of both schemes altogether, namely, low overhead for decoding with the H scheme, and high embedding capacity with the DE scheme, and take the characteristics of original image, that is, the smooth and active regions, into account to add the flexibility for reversible data hiding. Simulation results with our algorithm have demonstrated its applicability and the out performance over existing schemes.