Reversible multi-watermarking for color images with grayscale invariance

As a special way of hiding information, reversible data hiding is mostly used to embed data into digital multimedia. The original multimedia and embedded data can be restored from the watermarked one without any loss. Being different from the traditional reversible data hiding methods, the gray scale-invariant reversible watermarking method, which keeps the grayscale of image unchanged as the information is embedded, is proposed for color images recently. Although color images are widely used in practice, there are more reversible data hiding algorithms and feature points extraction algorithms for gray images rather than color images. In this paper, two multiple watermarking mechanisms have been proposed for color images with grayscale invariance, the multi-level watermarking mechanism, where one feature region is selected and the watermarks are embedded for multiple times, and the multi-region watermarking mechanism, where the multiple non-overlapping feature regions are selected to embed watermarks. Different from others, the former mechanism uses multiple embeddings based on the feature regions to increase the embedding capacity and the latter one uses local embedding instead of global embedding to reduce the impact on the whole image. At the same time, the selection of feature points can meet certain conditions and get more suitable regions for information embedding. Experimental results show that the proposed scheme can extend the capacity efficiently while keep the characteristic of grayscale invariance.