Research On Face Recognition Based On Fusion Of Infrared And Visible Light

The current mainstream face recognition systems are based on the integrated processing of visible light face images mostly,and its recognition effects are easily disturbed by factors such as brightness and angle of the ambient light.While the infrared face images can not only effectively avoid this adverse effect,but also be obtained without visible light,even at night,so this paper considers the fusion processing to improve the overall quality of the face image.For digital images,the local information content is commonly characterized by the corresponding pixels and their neighborhood.Therefore,in this paper,several groups of face images are selected randomly in the USTC-NVIE database and wavelet transforms are performed respectively.The low frequency threshold equalization and high frequency window energy method fusion rules are proposed for high and low frequency coefficient region characteristics.In the process of face image preprocessing,a multi-scale cascading detection method based on OpenCV is used to perform human eye positioning,and to intercept the effective region of the human face,and the performing scale normalization on the infrared and visible light face images respectively.In the evaluation of the final fusion face image quality,this paper proposes a quantitative synthesis analysis method for global assessment.The experimental results show that this fusion scheme can Increase the comprehensive characterization information of high-quality infrared and visible light face images by more than 5 percent,and this fusion scheme is effective for improving the overall quality of face images and it is worth researching further.