Kinship Verification Analysis Based on Color Features
Nermeen Nader, Fatma El-Zahraa A. El-Gamal, Mohammed Mahfouz Elmogy · 2022
Kinship verification is an attractive research area as it has been used in several applications, such as finding missing relatives and family album analysis. This paper proposes a new technique for verifying kin relationship dependent on a new color feature (color moments). It consists of 4 stages. First, preprocessing stage is responsible for improving the quality of input images by transforming the image from the red, green, and blue (RGB) color space into a lightness, channel A and channel B (LAB) color model. Second, the feature extraction stage is responsible for extracting the most discriminative information. The third is the pair representation stage, which combines the extracted feature vectors into one vector. Finally, the stage responsible for decision whether the two input images are kin or not, namely, kinship verification. The evaluated mean accuracy on KinFaceW-II was 83.9%. Therefore, the proposed method’s mean accuracy outperforms many of the state-of-the-art approaches, as well as convolution neural networks (CNN).