Automatic verification of kinship based on facial images

Abstract

Facial Kinship Verification, the process of identifying family relationships through facial images, is gaining more interest due to its many applications. This work proposes the use of feature extraction technique for kinship verification: the Histogram of Local Binary Patterns (Hist LBP). Hist LBP offers a richer representation of facial features by capturing both texture and color information. We evaluate its effectiveness against established methods like Local Phase Quantization (LPQ) and Local Binary Patterns (LBP) across various color spaces to assess its potential for improvement. Furthermore, we explore the impact of different normalization techniques applied to the Hist LBP descriptor in order to further enhance performance. The normalized Hist LBP features are then fed into a Support Vector Machine (SVM) classifier with 5-fold cross-validation for robust model training and evaluation.We evaluate the system’s performance using the KinFaceW-II benchmark, allowing us to analyze the impact of Hist LBP on kinship verification accuracy compared to existing approaches. Ultimately, this work strives to contribute to the advancement of facial kinship verification technology by providing a robust and effective solution.