基于预测误差扩展的高动态范围图像可逆数据隐藏

Kinds of reversible data hiding methods for low dynamic range (LDR) have been discussed. However, there is less research about high dynamic range (HDR). In this paper, a reversible data hiding algorihtm for HDR iamges based on prediction-error expansion (PEE) is proposed. Firstly, HDR images are encoded by radiance format. Secondly, the redundancy existing among channels is employed to classify the pixels into flat and edge areas. Thirdly, two differnet embedding manners are adopted for different areas for a high capacity. Besides, to further reduce the visual distortion, an embedding order is proposed to make sure that data is always hidden into pixels with less distortion. Finally, after message extraction original HDR images can be exactly recovered. Various experiment results show that compared with tradional methods, as more redundancy among color channels are exploited, the proposed method can achieve a hig embedding capacity with low distortion in the marked images.