Table 4.
Characteristics of the image-transformation-based feature extraction methodologies.
| Ref. | Key Features | Advantages | Disadvantages |
|---|---|---|---|
| [76] | Multi-scale self-adaptive enhancement transformation | Very fast, a low EER (0.13%) | Timing performance is not reported |
| [77] | Usage of the Radon transformation for driver identification | High accuracy rate (99.2%) for personal identification |
Tested upon a small dataset |
| [80] | Embedded system using the HAAR classifier | Fast recognition time and low computational complexity | Accuracy analysis is not reported |
| [81] | Second generation of wavelet transformation | Fast, a low EER (0.07%) | Dataset and experimental information are missing |
| [82] | Combination of the Radon transformation and common spatial patterns | Fast, a high RR (100%) | Small dataset |
| [83] | Usage of Discrete Wavelet Packet Transform decomposition at every sub-band | Improves upon Discrete Wavelet Transform and the original DWPT | Low RR (92.33%) |
| [84] | Variable-scale USSFT coefficients | High reliability against blurred images | Low RR (91.89%) |
| [85] | Usage of the Haar Wavelet Transformation | High accuracy (99.80%) | Accuracy highly depends on parameters |
| [86] | Feature enhancement and extraction using the Radon transformation | Improvement in accuracy in contacted and contactless databases | High EER (1.03%) |
| [87] | Usage of adaptive vector field estimation using spatial curve filters through effective curve length field estimation | Low EER (0.20%), improves recognition performance compared with other methods | Performance analysis is missing |
| [88] | Usage of Discrete Wavelet Transform | A hardware device is proposed | Small dataset |
| [89] | Fusion of the Hilbert–Hung, Radon, and Dual-Tree wavelet transformations | Low EER (0.014%) and improves upon other methods | Three vein images from different parts |