Improved Weighted Average-based Rhombus Predictor in Reversible Data Hiding Using Prediction Error Expansion
Ling Chu, Haowen Wu, Yilin Zeng, Xin Lai Tang · 2021 IEEE 6th International Conference on Computer and Communication Systems (ICCCS) · 2021
Rhombus predictor is a popular way to achieve prediction error expansion based reversible watermarking algorithm because of its superior performance. However, most of the detailed implementations were designed only for images with a majority of smooth regions leading to large prediction errors by using directly on texture regions. Despite the improvements, they failed to adequately consider correlations between pixels, moreover, they used inferior fluctuation-based sorting mechanisms. To cope with such problems, we designed a new weight calculation technique for the rhombus predictor. Specifically, based on the complexities of texture regions evaluated by correlations of four neighboring pixels, we ensure that texture regions have always smaller weights. Furthermore, to guarantee higher imperceptibility, we propose a new fluctuation-based sorting mechanism. Experimental results indicate that our predictor achieves lower distortion while obtaining high embedding capacity in comparison to the state-of-the-art.