Reversible Image Data Hiding Using Quad-tree Segmentation and Histogram Shifting

This paper is working towards a high capacity histogram-based reversible data hiding algorithm with a relatively lower distortion introduced after embedding the secret message. The proposed algorithm can be thought of as an improved version of the traditional histogram shifting method. To take advantage of simple local distribution of pixel intensities, the idea behind the proposed algorithm is to consider the multiple local histograms, each of which is derived from pixels in a locally contiguous sub-image area, instead of the single histogram for the whole image area. We develop a hierarchical segmentation scheme that is able to hierarchically partition the input host image into several variable-sized blocks of pixels under a maximal capacity criterion for each block partition. These partitioned blocks are organized as a tree structure for the ease of representation. The secret message and the partition tree information are then embedded in these image blocks. With the proposed segmentation scheme, the algorithm can easily find a suitable non-overlapped partition of the image to significantly increase embedding capacity. Experimental results reveal that the proposed method can indeed provide higher embedding capacity than the traditional histogram shifting method while maintaining a high visual quality that is comparable to the traditional one without spatial segmentation scheme.