Robust color image watermarking algorithm based on synchronization correction with multi-layer perceptron and Cauchy distribution model

A robust color image watermarking algorithm based on the multi-layer Perceptron (MLP) and Cauchy distribution model is proposed. The algorithm has great performance against geometric attacks. Firstly, the central region of the Y-color channel is extracted, and the watermark is embedded into the low-frequency component by dual-tree complex wavelet transform (DTCWT). A synchronization correction technique based on the MLP model is used prior to the watermark extraction. When the MLP model is trained, the position and scale parameters of the Cauchy distribution are employed to generate the image feature vector after first calculating the quaternion discrete cosine transform (QDCT) of the original color host image. The new 3D hyperchaotic mapping is used during the algorithmic process to increase security, which is employed for scrambling, encryption, and embedding of watermarks, respectively. The experimental results demonstrate that the proposed algorithm has good imperceptibility, accurate model prediction, and excellent robustness against common image processing attacks and geometric attacks.