Enhanced biometrics face morph detection via neural style transfer and 2D wavelet transformation

Biometric feature identification has grown in popularity during the past a decade. Nowadays, many mobile devices are fitted with common components such as fingerprint sensors and high-quality video, which have grown more affordable. Even systems requiring strict security regulations, such as border control at airports, have turned to biometric technologies due to their advanced precision in facial recognition. This study proposes an approach to detection based on 2D Wavelet Transformation and Progressive Enhanced Learning Framework (PELF). Additionally, using the neural morph enhancement process, to study Deep Learning-based Neural Style Transfer (Deep-NST) model was developed to identify morphing face images with features resembling those of the input images. To precisely capture the features and texture changes in the image, our approach first extracts high-frequency information from all three color channels. Next, a progressive enhanced learning framework was built to combine high-frequency data with RGB information. Furthermore, the framework that is suggested for assessing the quality of our method and contrasting it with existing DNN interpretability techniques. Deep-NST is superior to other methods in identifying abnormalities. When the DNN provides an incorrect or confused evaluation, this becomes extremely crucial.