Detection of Signature Forgery Using Morphological Features: A Review of Structural Characteristics and Statistical Approaches

The traditional role of handwritten signatures has been in uniquely identify, authorise, and confirm transactions within financial, legal, administrative and forensic domains. In an ever-increasing digitised world of data entry, disputed signatures are frequently incorporated into questioned document examination. Traditional forensic inspection methods rely on subjective evaluation of writing styles by experts, but modern analysis methods involve increasing quantification used to distinguish authenticated from forged signatures. Among these quantitative analysis methods is morphologic analysis, a relatively straightforward method using measurable structural and geometric properties derived from images of signatures. Various features associated with writing height, width, aspect ratio, area, perimeter, eccentricity, density of pixels, contour shape, distribution of pixels, stroke and structural elements and spatial relationships can provide a good indication of the morphology of signature. The article reviews the relevance of morphological features in handwritten signature forgery detection using offline signature verification as a focal theme. Information within the literature from geometric, structural, directional, graphometric, texture and machine learning approaches is reviewed. The use of statistical methodology in signature verification research is explored, including descriptivestatistics, hypothesis testing, Correlation Analysis, Principal Component Analysis, Fisher Linear Discriminant Analysis, Support Vector Machine (SVM), Neural Network (NN), and performance indices like Accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR) and Error Equality Rate (ERR). Comparing key studies, it was revealedthat noneof the single features can authenticate a signature; usually, a combination of features yields effective distinction. Key limiting factors such as the natural variation of a writer's signature (intra-writer variance), image acquisition condition variations, scarcity of forensic signature databases, possibility of high-skilled forgeries, class-imbalanced data, and lack of unified reporting standards for these analyses were considered. More importantly, it requires correlation of both interpretable morphologic values with rigorous statistical modelling, Explainable Artificial Intelligence, standard forensic datasets, and a forensic validation approach.