Table 1 The summary of related works.
From: Convolutional neural network and wavelet composite against geometric attacks a watermarking approach
Reference | Technique | Strengths | Limitations |
|---|---|---|---|
Zear et al.23 | DWT + DCT + SVD | Multiple watermarking, good robustness | Moderate imperceptibility, limited payload |
Fazli & Moeini24 | DWT + DCT + SVD | Geometric attack resistance | Complex embedding, limited scalability |
Singh et al.25 | SVD + BPNN | Noise reduction via neural networks | Lower PSNR (~ 34.88 dB), limited robustness |
Ernawan & Ariatmanto27 | Scaling factor techniques | Copyright protection, flexible scaling | Vulnerable to geometric distortions |
Awasthi & Srivastava28 | SURF + LWT + DCT + SVD | Efficient descriptor computation | High computational cost |
Ravichandran et al.30 | ROI-based watermarking | Accurate tamper detection | Application-specific, limited generalizability |
AlShaikh et al.31 | Pigeon algorithm + DCT | Optimised embedding locations | Limited to DCT domain |
Kumar et al.32 | PSO + LWT + Hessenberg | Robust medical watermarking | Requires tuning of PSO parameters |
Begum et al.39 | DWT + SVD | Enhanced imperceptibility and robustness | Vulnerable to false positives in SVD |
Tang & Zhou40 | DWT + SVD + FRFT | Fractional domain robustness | Complex transform domain operations |
Adi et al.42 | Adaptive matrix + chaotic sequencing | Efficient tamper detection | Fragile watermarking, less robust |
Mali & Agilandeeswari43 | DCNN + wavelet + fractional optimization | Deep learning-based robustness | High training time, complex architecture |
Proposed work | DWT + SVD + CNN + Compression | High payload, imperceptibility, robustness, and multi-watermarking | Requires CNN training, moderate complexity |
