Image hiding algorithm based on local binary pattern and compressive sensing

Abstract

In recent years, with the rapid growth of digital communication, the protection of image privacy has become a critical concern. Traditional image encryption methods may attract attacker attention due to the noise-like appearance of cipher image. To address the theft risk to the private image, a novel three-dimensional chaotic map of Chebyshev coupled Logistic with Sine map (3D-CCLSM) is designed, and then an image hiding algorithm based on local binary pattern (LBP) and compressive sensing (CS) is proposed, named ImHALC. By integrating LBP-based texture feature extraction and CS-based image compression, ImHALC aims to enhance both the security and imperceptibility for steganographic image. Especially, LBP is taken to connect the plain image and keystream, resulting a high security for ImHALC. Firstly, in stage of keystream generation, texture information of the plain image is extracted by using LBP and seen as input of hash function SHA-256, to produce corresponding hash values. Then, these hash values are used to generate the initial value of 3D-CCLSM by a new designed key transformation model (KAM), so as to get the keystream for encryption. Secondly, in stage of image compression, a measurement matrix is constructed by above keystream, and CS is applied to the plain image to get measurements. Thirdly, in stage of image encryption, measurements are confused and diffused to produce a cipher image. Finally, in stage of embedding, an embedding method using integer wavelet transformation (IWT) and 2^k correction is presented, so as to embed the secrets (i.e., cipher image) into a given carrier image to obtain hiding performance, i.e., forming a carrier image hiding secrets (CHS). In particular, the ImHALC can achieve the effect of blind extraction. After adopting a two-dimensional projection gradient algorithm with embedded decryption (2DPG-ED), the reconstruction quality for the plain image is good for test images.