Research On Face Recognition In Complex Illumination

With the rapid development of modern face recognition technology,face recognition technology has shown great application prospects in many areas such as state and public safety.At present,face recognition technology has become an important part of machine vision.Although face recognition technology has been widely applied in many fields,there are still many problems in face recognition research.In the current technology development of face recognition,there are many factors that affect the technology of face recognition.Illumination is an important factor.In this paper,we will focus on analyzing the face recognition algorithm under complex illumination conditions in face database.The main work of this article is as follows:(1)In this paper,an adaptive Gamma correction method based on multi threshold OTSU segmentation is proposed.Through the experimental analysis and comparison,it is found that the method can adaptively divide the image region into dark,transition and bright intervals,and can carry out different Gamma corrections in each interval.In this algorithm,the method of OTSU segmentation is introduced to solve the problem of block effect in traditional segmentation,and the multi threshold method avoids the problem of traversing image pixels.The experimental results finally verify that the method has good processing effect in improving the image light,enhancing the quality of the image and improving the visual effect of the image.(2)In this paper,proposes a face image enhancement algorithm based on Lambert illumination model: multiscale Retinex algorithm(MSR),self quotient image algorithm(SQI),and wavelet transform based processing method(WT).The effectiveness of the algorithm is verified by experiments,and the following conclusions are obtained: MSR is easy to have a "halo" phenomenon when dealing with light.In the SQI method,different window size and filter core scale have different effects on the processing effect,and a better processing effect can be obtained by using the multiscale SQI method;WT method It can achieve better illumination effect,but the processed image will appear blocking phenomenon,and we can choose orthogonal and smooth wavelet basis to improve this situation.(3)In this paper,a face recognition algorithm based on Curvelet transform and Retinex theory is proposed.The algorithm uses the good edge preserving ability of Curvelet transform as a tool for multi-scale decomposition,and uses bilateral transformation and threshold denoising to deal with the low frequency and high frequency coefficients respectively.Finally,the method is obtained by the Retinex theory method.The light is characterized by constant illumination.By comparing with the experiments of SSR,MSR,SQI and WT,it can be seen that the algorithmcan make full use of the features that can be used for recognition in the high and low frequency coefficients of the image,and can achieve better recognition effect under the condition of changing light.