Sparse Image Reconstruction by employing Adaptive Gradient Algorithm in Image Steganography

Abstract

Image steganography is a well-known technique to transmit a secret message in the form of image from one place to another without giving even a small sense to the third party. Stego attack is a common problem in the field of image steganography. However secure is the key and the algorithm, the third party may attack/jam the secret message which is being communicated. This paper deals with solution to such stego attack problem by involving two levels of security and by employing sparse based image reconstruction technique to retrieve the jammed/corrupted secret message. In this paper, two models for image steganography is proposed by involving adaptive gradient algorithm for secret image reconstruction. The stego attacked/corrupted secret image can be completely reconstructed from its very few non-corrupted image pixels by adaptively estimating the gradient of the error function with respect to the pixels to be estimated subjected to l₁-norm for including sparsity in the transform domain of the image. Results obtained from the simulation shows that this proposed method performs better with respect to peak signal to noise ratio and computation time as compared with other conventional image filtering techniques.