Robust image watermarking based on multiwavelet parent-child relation and adaptive support vector regression

In this paper, a robust image watermarking algorithm in multiwavelet domain is proposed. Multiwavelet transform provides good energy compaction and trade off between imperceptibility and watermark robustness properly. In the proposed algorithm, after performing a three-level multiwavelet decomposition, a parent-child relationship is defined between coefficients in different subbands. To increase the robustness, the relationship between the root node and its offspring nodes is learnt by support vector regression (SVR) and then the watermark is embedded and extracted based on this inherent relation. The good energy compaction property of multiwavelet can cause small number of coefficients containing most of image energy, therefore the range of target output values of SVR is extended which notably made inaccuracy in predictions. In order to overcome this problem, we proposed a procedure, which applies the inputs to SVR in an adaptive way. The experimental results show, the proposed algorithm possesses acceptable perceptual quality, while it is robust against common attacks such as JPEG, low-pass filtering, noise addition, cropping and resizing, etc.