A secure image steganography based on JND model

Abstract

Minimizing distortion produced by embedding process is very important to improve the security of hidden message and maintain the high visual quality of stego images. To achieve these objectives, an effective strategy is to perform pixel selection which is well-known as a channel selection rule. In this approach, a pixel associated with the smallest image degradation is chosen to carry secret bits. From these facts, in this paper, a new secure channel selection rule for digital images in spatial domain is designed and proposed. In this new approach, the modified matrix embedding method is utilized as data hiding method because it introduces more than one embedding change to be performed. This enables us to select a suitable pixel to embed message bits with less degradation yielded in a stego-image. In pixel selection of the proposed method, a just noticeable difference value and gradient value of a considering pixel are employed together. The experimental results (which were conducted on 10,000 uncompressed images) indicate that stego images of the proposed approach achieve a higher perceptual quality and security than those of the stego-images created by the previous approaches.