High-Capacity Constructive Steganography Using Optimal Texture Block Synthesis

Abstract

In this paper, we design a new high-capacity constructive steganographic scheme by optimal texture image block synthesis. First, original texture pattern is divided into multiple seed patches, which are hashed into a blank stego pattern by the key. Furthermore, we employ generated seed patches to reconstruct a candidate pattern, which is scanned by sliding window to generate a series of candidate texture blocks. Subsequently, we sort these candidate texture blocks and select the appropriate ones that their serial number equals to the secret messages to achieve information hiding. After theses texture blocks containing secret messages are hashed into the stego pattern by the key, the remaining positions of stego pattern are sequentially filled with the optimal candidate texture patches by calculating the least square error. Finally, the stego pattern is completely synthesized as the constructive stego image. Compared with the existing schemes, our proposed scheme has a significant high embedding capacity, while keeping a higher visual quality of stego images.