Table 1 Summary of existing techniques in the literature.
From: A hybrid machine learning framework for offline signature verification using gray wolf optimization
Ref. No. | Authors | Proposed Technique | Advantages |
|---|---|---|---|
Ojala et al. | LBP for texture classification. | Pioneering work on rotation-invariant texture classification. | |
Pal et al. | LBP and ULBP for Hindi/Bangla signatures. | Explored OSV for non-Latin scripts. Provided a baseline for LBP/ULBP performance. | |
Ferrer et al. | LBP and LDP for binary signatures. | Identified LDP as effective for black-and-white signature images. | |
Guerbai et al. | A variant of OC-SVM with combined distance metrics. | Addressed the small sample size problem common in real-world OSV. | |
Yilmaz et al. | Local Histogram features combining HOG and LBP per image region. | Used combined local features for a robust representation. Did not require skilled forgeries for training. | |
Serdouk et al. | Fusion of OC-LBP and LRFs. | Proposed a novel fused feature. Achieved a low AER reduction and reduced dataset size. | |
Kiani et al. | Radon Transform for feature extraction, classified with SVM. | Achieved a very low FRR of 2%. | |
Zulkarnain et al. | Triangulation geometric features with a Voting-based classifier. | Introduced a novel geometric feature set derived from the signature’s center of gravity. | |
Panchal et al. | Shape-based geometric features (area, eccentricity) with an ANN classifier. | Utilized simple, computationally inexpensive features. | |
Pandya | Hierarchical clustering technique for verification. | Employed a simple clustering approach. | |
Engin et al. | CNN model for noisy documents. | Achieved state-of-the-art accuracy on real-world noisy documents. | |
Bertolini et al. | Comparative analysis of LBP and LPQ for writer identification. | Direct comparison of texture descriptors for a related biometric task. | |
Singh et al. | Division of handwriting into texture blocks for LBP and CSLBCoP. | Detailed block-based analysis using advanced descriptors. | |
Bahram | Combination of MLBP and IWSL for writer identification. | Combined features to capture different aspects of writing style. | |
Fadaei et al. | LBP on sub-regions with feature optimization using PSO. | Used PSO for feature optimization. | |
Ahlawat et al. | Integration of LBP and a novel Octave Pattern for signature verification. | Enhanced both texture-based and structural signature characteristics. | |
Fadaei et al. | ScLBP for image retrieval. | Scale-invariant method validated on 7 diverse benchmark datasets. | |
Mashhadani et al. | Fusion of a Type-2 Neutrosophic Similarity Measure. | Directly addresses uncertainty, leading to higher accuracy and improved FAR/FRR over Type-1. | |
Abdulhussien et al. | Multi-feature fusion with GA-based selection and one-class learning. | Addresses Arabic script challenges and data imbalance. Validated on multiple datasets with a 5% improvement. |
