Blind Color Image Watermarking Using Deep Artificial Neural Network Using Statistical Features

With increasing digital content over the internet it is very important to secure the digital contents in such a way that the identity and integrity of data is preserved in some way and with lots of advancements in these fields, it is very important to protect digital contents such as healthcare data and electronic records. It is also significant to protect the digital contents for a smart city scenario where mostly records or data floats electronically. This paper presents a methodology for copyright protection and authorship of digital contents for smart city. In order to achieve a balance between imperceptibility and robustness a robust watermarking scheme is proposed using deep artificial neural network (DNN) and lifting wavelet transform in YIQ color domain. YIQ color model is utilized for digital health image decomposition. Statistical features have been obtained for creating training and testing set from the healthcare datasets. Here watermark extraction is done as a binary classification technique and PCA is utilized for reducing the feature set. Ten standard images have been used for image watermarking and for threshold value 0.3 it shows the average imperceptibility of 51.08 dB and shows good robustness under various image attacks.