LAUSR

This paper proposes a new image steganographic method that effectively combines LSB embedding with Octa-PVD embedding. A cover image is divided into non-overlapping 3 × 3 sub-blocks, the n least significant bits of the center pixel of each sub-block is first substituted by secret data (n-LSB substitution). Then, the differences between the center pixel and its eight neighbors are calculated. For each direction, if the difference is equal or larger than a threshold (predefined by users or automatically determined by image analysis), secret data is embedded into the neighbor pixel by n-LSB substitution. Otherwise, secret data is embedded by PVD embedding, but into the neighbor pixel only. Consequently, depending on the conditions of each sub-block, a single embedding method can be used to the whole sub-block, or two embedding methods can be used alternately within a sub-block. Comparisons with existing LSB or multi-directional PVD embedding methods demonstrate that the proposed method has more optimized and higher embedding capacity and PSNR.