Distortion function for steganography in texture synthesized images

Abstract

This paper proposes a distortion function for steganography in texture synthesized images. Given a small piece of texture, an image synthesis algorithm is employed to generate a texture image in arbitrary size with similar local appearance. The obtained texture image is used as cover for data embedding. A distortion function is designed for the cover image to measure the detection risk of modifications. The image texture, splicing of patches, and repetition of texture blocks are contained in the proposed distortion function to fit the properties of synthesized images, which results in high undetectability against steganalysis. Experimental results also prove that the proposed distortion function performs better than current state-of-the-art steganographic methods.