Edge detection based JND model for digital watermarking

In the watermarking schemes, Just Noticeable Distortion (JND) model provides an efficient method to keep the balance between watermark's imperceptibility and robustness. Traditional JND models are not as perfect as they should be. Edge detection, which extracts content features well, improves the performance of JND models. In this paper we explore the reasons and simulate JND models based on edge detection either in spatial domain or in DCT domain. The experiment results show that JND models based on edge detection in DCT domain have better performance than those in spatial domain.