An image embedding in image by a complexity based regionsegmentation method

摘要

This paper describes a new technique to embed secret data into adummy image by using image segmentation based on a local complexitymeasure. The key idea to this approach is that a binary image can becategorized as “informative” and “noise-like”regions which are segmented by a “complexity measure”. Ifthe embedding data is noise-like, we can hide it in the noise-likeregion of the dummy image. If a part of embedding data is simple, thenwe apply “image conjugate” operation to it. This operationtransform a simple pattern into a complex pattern. In our experiment, wecould embed two color images in a 512×512 (8 bits/pixel) size grayimage (which was dummy) without losing any information. The total amountof the two embedded images was 115 KB, which was about 45% of the dummyimage