Table 8 Comparisons of features among different schemes
From: Coverless image steganography using morphed face recognition based on convolutional neural network
Schemes | Coverless | Methodology | Capacity | Scalability in dataset | Resist steganalysis |
|---|---|---|---|---|---|
Chang et al.’s scheme [6] | No | Pixels value modification | Very high | None | Difficult |
Wang et al.’s scheme [5] |  | Pixels value modification | Very high | None | Difficult |
Chang et al.’s scheme [9] |  | Pixels value modification | Very high | None | Difficult |
Zhou et al.’s Scheme [27] | Yes | Feature based mapping (Hash) | Low (log2N) | No | Easy |
Yuan et al.’s Scheme [29] |  | Feature based mapping (Hash) | Low (log2N) | No | Easy |
Zhang et al.’s Scheme [3] |  | Feature based mapping (DCT) | Low (log2N) | No | Easy |
Mu and Zhou’s scheme [33] |  | Feature based mapping | Low | No | Easy |
Liu et al.’s scheme [34] |  | Feature based mapping (DenseNet) | Low (log2N) | No | Easy |
Proposed scheme |  | Mapping by morphing and recognition (MFR-Net) | High (log2((N2 + N)/2)) | Yes | Easy |