Abstract
Generalizations of the pixel-value differencing (PVD) method for data hiding in graylevel images are proposed. Two extensions of the PVD method are analyzed: the block-based approach and the Haar-based approach. For the block-based PVD, the cover image is divided into non-overlapping horizontal or square blocks of n pixels. In each block, n − 1 differences are calculated between consecutive pixels. These differences are classified to embed the secret message. For the Haar-based PVD, the 2-D integer Haar wavelet is applied to decomposed the cover image. The high-frequency components are used to hide message. Higher level of Haar decomposition allows more bits to be embedded. For both proposed generalizations, the capacity of the embeddedmessage is significantly increased. Moreover, both generalizations are invulnerable to the RS-diagram and histogram steganalysis.