Skin Segmentation-Based Disguised Face Recognition Using Deep Learning

Abstract

Disguised face recognition refers to the ability of a computer system or algorithm to identify a person’s face even when they are wearing some form of disguise, such as a mask, hat, sunglasses, or makeup. This is a challenging problem in computer vision and pattern recognition, as disguises can significantly alter a person’s facial features and appearance, making it difficult to match the image with a known face. In this work, we propose a novel approach to disguised face recognition by focusing on the skin regions of the face, which are less likely to be covered by disguises. Skin regions are segmented from the faces by applying a region-based marker-controlled watershed algorithm and features are extracted from these skin regions using a convolutional neural network and classify the disguised faces using a deep learning-based recognition model. The results show that the proposed model achieves high accuracy on $64\times 64$ image size, with an overall accuracy of 94.92%. We also performed an ablation study to analyze the impact of different factors on the performance of the proposed approach, including the image size and the size of the kernel filter. Overall, our approach provides a promising solution for the challenging problem of disguised face recognition.

Keywords

• Deep learning
• disguised faces
• skin segmentation
• Sejong face dataset
• region-based marker-controlled watershed algorithm