Infrared face image conversion method based on DCGAN

Abstract

The invention discloses an infrared face image conversion method based on DCGAN, and the method comprises the steps: building a DCGAN confrontation generation network, and the DCGAN confrontation generation network comprising a generator and a discriminator; and inputting the visible face image and the near-infrared face image of the training set into a generator of a DCGAN confrontation generation network, performing image generation by adding a noise signal, performing confrontation training on the generator by a discriminator, and finally obtaining the generator capable of generating the near-infrared face image through the visible face image. According to the method, the DCGAN network is utilized to convert the visible face image into the near-infrared face image, so that the face image has certain robustness on illumination and other influence factors, the difficulty of heterogeneous face recognition is reduced, and the recognition accuracy of heterogeneous face recognition is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a DCGAN-based infrared face image conversion method. Background technique [0002] Face recognition has been a rapidly developing research field for many years, and it is currently a research hotspot in the field of pattern recognition and computer vision. Facial recognition identifies or verifies identities by comparing faces to images of faces stored in databases. Near-infrared (NIR) imaging is more robust to illumination changes because it captures emitted energy from objects to produce images. Therefore, the use of near-infrared images has significant advantages over visible images for facial recognition under different lighting conditions. In real life, everyone has at least one visible light face image, such as an ID card photo, where the lighting conditions are controllable and there is no large expression. When face recognition is applied in places such as sec...

AI Technical Summary

Problems solved by technology

When face recognition is applied in places such as security check stations, in order not to be affected by conditions such as lighting, near-infrared images are used for face recognition, resulting in the problem of heterogeneous face recognition

In recognition, in order to improve the accuracy, the visible face image can be converted into an infrared image for subsequent recognition. However, the existing image conversion processing technology greatly increases the difficulty of heterogeneous face recognition, thereby reducing the recognition accuracy

Images