A blind image detection method for information hiding with double random-phase encoding

In this paper, a blind image detection method based on a statistical hypothesis test for information hiding with double random-phase encoding (DRPE) is proposed. This method aims to establish a quantitative criterion which is used to judge whether there is secret information embedded in the detected image. The main process can be described as follows: at the beginning, we decompose the detected gray-scale image into 8 bit planes considering it has 256 gray levels, and suppose that a secret image has been hidden in the detected image after it was encrypted by DRPE, thus the lower bit planes of the detected image exhibit strong randomness. Then, we divide the bit plane to be tested into many windows, and establish a statistical variable to measure the relativity between pixels in every window. Finally, judge whether the secret image exists in the detected image by operating the t test on all statistical variables. Numerical simulation shows that the accuracy is quite satisfactory, when we need to distinguish the images carrying secret information from a large amount of images.