An Information Hiding Method Based on Grouping Color Palette Indices by K-means Clustering and Similar Pairs
Chyuan-Huei Thomas Yang, Yu-Wei Hsu, Che-Lun Chun · 2012
We propose a data hiding method with color image palette. Many previous authors often embed data into the index table of the color palette or into the color palette directly. Those data hiding methods will change the color palette into a difference one. It is impossible to reconstruct the original color palette. We follow previous method to apply the K-means clustering to split the color image palette into several groups first. If a palette group has k numbers indices, the data can be embedded ¾ log2k Î bits into the pixel that its palette index falls in this group. the current embedding pixel will be replaced by the same group of pixel in the order of embedding data value which are corresponding each palette index. Next step, we collect those k {2 ¾ log2k Î in each groups to use the Euclidean distance to cluster two groups with similar group that have similar index pairs. We embed one bit in each pixel that belongs these two groups. the extraction method firstly groups of the color palette index of stego-image by k-means clustering and similar pairs like embedding steps. If the palette index of the pixel belong to groups that produced by K-means clustering, then finds what order in its group. That order is the embedded value. If the palette index of the pixel belong to groups that yielded by similar pairs, then it belongs to first group is embedded 0, else is 1. the information can be extracted completely until all embedded pixels are retrieved or retrieved to the last bit of host image. from the experimental results, the method has the more embedding capacity then it only uses k-means clustering in the same image.