Research And Implementation Of Low Illumination Face Image Recognition Method

With the rapid development of computer vision technology,face recognition systems and computer vision monitoring equipment can capture a large amount of image information.However,in low-light environments such as low-light indoors or unevenly-lit outdoor environments,due to the insufficient illumination of unnatural light sources,the reflected light on the target surface is weak,resulting in insufficient light passing into the imaging sensor,resulting in low The quality of the face image obtained under the illumination environment is severely degraded,and the image recognition performance is poor.It is difficult to discern the details in the image,which reduces the recognition rate of the face recognition system.Most of the traditional image enhancement algorithms enhance the image from the spatial domain and the frequency domain.In view of this situation,this paper from the perspective of image enhancement theory and method in low illumination environment,makes an in-depth study on the brightness,detail information and color restoration of the enhanced image.Firstly,this paper makes an in-depth analysis of the principle and advantages and disadvantages of traditional low-illumination image enhancement algorithms,including histogram equalization based enhancement methods,homomorphic resolution enhancement methods and the widely used surround Retinex method.Secondly,in view of the shortcomings of excessive enhancement,detail blurring and color distortion around Retinex method in the surround Retinex method,This paper proposes a multi-scale Retinex low-illumination image enhancement algorithm based on guided filtering.In the HSI color space,the saturation component is processed by linear stretching to increase the image color saturation and enrich the color;At the same time,the proposed algorithm performs non-linear global brightness correction on the illuminance component of the image,which improves the contrast of the image;Then,a multi-scale Retinex algorithm based on guided filtering is designed to enhance the corrected luminance component;Finally,the enhanced brightness component,saturation component and hue component are restored from the HSI color space to the RGB color space and processed for color restoration,finally,an enhanced image is obtained.The paper chooses the Ada Boost function in Open CV as the platform's face detection algorithm,and the Eigen Face method is selected as the face recognition algorithm in the platform.Combined with the proposed multi-scale Retinex algorithm based on guided filtering,the face recognition platform is constructed.Through the test,the platform can effectively recognize the low-illumination face image,and the recognition accuracy is greatly improved compared with other image enhancement methods.