A screen-shooting resilient data-hiding algorithm based on two-level singular value decomposition

Abstract

This paper presents a screen-shooting resilient data-hiding scheme for JPEG images based on a two-level Singular Value Decomposition (SVD). The mid-to-high frequency coefficients are chosen to balance the data-embedding robustness and the image's visual quality. At the same time, the two-level singular value decomposition (SVD) is additionally utilized to construct the feature matrices, thereby enhancing the robustness of the embedded message. The combination of DCT and two-level SVD enables achieving a high-performance screen-shooting resilient data-hiding algorithm. Moreover, the affine transformation is employed to rectify the distortion that caused by screen-shooting, thus further improving the data extraction accuracy. Experimental results demonstrate a minimum 3 % improvement in confidential information extraction accuracy, with the data extraction rate exceeding 92 % across screen-shooting angles ranging from −40 to 40°. The performance of the proposed algorithm outperforms other state-of-the-art schemes apparently.