Steganography based on region segmentation with a complexity measure

This paper describes a new steganographic technique using image segmentation based on a local complexity measure. The key to this approach is that a binary image can be divided into “informative” and “noiselike” regions by using a “complexity measure.” This property allows us to embed secret data in the noiselike regions if the secret data have a random pattern. If the secret data are simple, then we apply an image conjugate operation that transforms a simple pattern into a complex pattern. In the experiment, a JPEG file was embedded into a gray scale dummy image (256 × 256, 8 bits/pixel) without losing any visual information. © 1999 Scripta Technica, Syst Comp Jpn, 30(3): 1–9, 1999