LAUSR

In this paper, we present curvelet transform (CT) based image steganography that embeds scrambled secret image in appropriately selected cover image. Curvelet transform offers optimal nonadaptive sparse representation of objects with edges and possesses high directional sensitivity and anisotropy. Cover image is decomposed using curvelet transform and adaptive block based embedding is carried out only in non-uniform regions of high frequency curvelet coefficients. In addition, this work also demonstrates a new cover selection method to choose suitable cover from image database. Spatial information based image complexity is modelled using fuzzy logic to identify set of images that yields least detectable stego image. From this set of ranked images, best cover can be chosen for carrying secret information depending on amount of information to be embedded. Cover selection offers reduced risk of detectability and ensures security. It is evident from experimental results that proposed method outperforms conventional methods in terms of imperceptibility, robustness and security.