Guided filtering based color image reversible data hiding

In this paper, a color-image-dedicated reversible data hiding (RDH) algorithm is proposed to improve embedding performance by applying a guided filtering predictor and an adaptive prediction-error expansion (PEE) scheme. PEE-based RDH methods can be mainly separated into two stages for each channel embedding, i.e., pixel prediction and prediction-error histogram (PEH) modification. In our work, the inter-channel correlation is exploited at all stages of prediction and modification. Specifically, for predicting the pixels in the current channel with the guidance of pixels from other channels, a linear transform model from reference channels to the current channel is established and its coefficients are determined by the Laplacian minimization criterion. Then, to modify the PEH, an adaptive PEE embedding scheme is conducted by seeking the optimal parameters of the embedding bins and the complexity threshold to minimize distortion. The experimental results demonstrate the proposed method has better performance than the state-of-the-art, color-image RDH methods.