Table 5.
Characteristics of the remaining feature extraction methodologies.
| Ref. | Key Features | Advantages | Disadvantages |
|---|---|---|---|
| [90] | Combination of morphological peak and valley detection | Precise details, better continuity compared with others, fast, and robust against noise | Low RR |
| [91] | Application of tri-value template fuzzy matching | Robust against fuzzy edges and tips, does not need correspondence among points, and has a low EER (0.54%) | A set of parameters needs optimization |
| [92] | Application of BLPOC | Simple preprocessing, fast with a low EER (0.98%) | A set of parameters needs optimization |
| [93] | Extraction of profile curve valley-shaped features | Fast, easy to implement, and satisfactory results | No classification results provided |
| [94] | Application of OPM | Enhances the similarity between samples in the same class | High EER (3.10%) |
| [95] | Application of PHGTOG | Reflects the global spatial layout and local gray, texture, and shape details and fast with a low EER (0.22%) | Personalized weights for each subject, a low RR (98.90%) |
| [96] | Feature code generation from a modified angle chain | Fast with a low EER (0.0582%) | Small dataset |
| [97] | Combination of a Frangi filter with the FAST and FREAK descriptors | Reliable structure and point-of-interest extraction | No classification results provided |
| [98] | Utilization of superpixel features | Extraction of high-level features | Requires setting of weights for the matching process, a high EER (1.47%) |
| [99] | Application of the Mandelbrot fractal model | Fast, a low EER (0.07%) | Dataset information is missing |
| [100] | Application of canny edge detection | Fast | Slow recognition time and a low RR |
| [101] | Application of Potential Energy Eigenvectors for recognition | Fast and higher accuracy compared with minutiae matching, a low EER (0.97%) | Not reported |
| [102] | Feature extraction using a SVM classifier | Consistent | Low accuracy rate (98.59%) |
| [103] | Feature contrast enhancement and affine transformation registration | Improved preprocessing, can reach a RR of 100% and an EER of 0% | Results vary highly |
| [104] | Combination of the SIFT and SURF keypoint descriptors | Robust to finger displacement and rotation | High EER (6.10%) and a low RR (93.9%) |
| [105] | Takes into account deformation via pixel-based 2D displacements | Low EER (0.40%) | Low timing performance |