Robust Color Image Watermarking using ANN and Statistical features for Safer Society
Manoj Kumar Pandey, Naresh Kumar Kar, Jyoti Upadhyay · 2025
Information security is a critical issue for both the nation and the smart city as it is also very important concept because it leads to the data security. To address this concern, a blind color image watermarking technique based on artificial neural networks (ANN) was proposed. This method involves the extraction of statistical features to construct a training and testing dataset sized at 512x14. Specifically, the blue color channel is chosen for embedding the watermark using quantization, and a 512-bit watermark is employed for experimental purposes. Watermark extraction is done using the classification problem and principal component analysis is used for reducing the feature set to 512x10 and proposed ANN based watermarking shows average robustness of 0.9074, 0.8216, 0.7028 and 0.8415 for Lena, Peppers, Mandril and Jet respectively and it also shows good imperceptibility of approx. 35dB for threshold value of 42. The proposed scheme shows good robustness against most of the attacks.