Steganography Based on High-Dimensional Reference Table

摘要

In this paper, a high dimensional reference table-based embedding is proposed for steganography. It employs an N-length cover vector to embed B(K, N)-ary notational digits, which causes a Euclidean distance-based distortion no more than K. By choosing the appropriate values of K and N, a embedding strategy suitable for arbitrary relative capacity can be achieved. Further, we introduce the One-by-One Search Algorithm (OOSA) to optimize the mapping function, which maps the elements in the codeword set to those in the neighborhood set. Due to the high dimensional embedding, the proposed scheme provides a higher embedding efficiency or relative capacity compared with other similar approaches. Experiments on both grayscale images and H.264 video show the effectiveness of the proposed scheme.