A Night Traffic Image Suppression Enhancement Algorithm Based On Zero Reference Curve Estimation

In order to better ensure the safety of pedestrians and road safety in nighttime environments,image enhancement for complex nighttime traffic scenes has always been a key research area for many scholars.However,due to interference factors such as severe noise,complex and variable light sources,and limited detail information in the sample images,the current enhancement algorithms are still not ideal.Therefore,this article summarizes and analyzes the shortcomings of relevant algorithms and proposes a traffic image suppression and enhancement algorithm suitable for complex environments with multiple light sources at night.The specific innovation points are as follows:(1)This paper proposes a subspace Attention De Glow(SSA-De Glow)image suppression network based on subspace attention to address the unique characteristics of night traffic scene images,such as complex and numerous light sources,severe effects of glow and glare,and strong noise.This network is used for image suppression and denoising to alleviate the problem of excessive enhancement of glow and glare in image enhancement.SSA-De Glow is based on the De Glow network,abandoning its complex iterative process and optimizing the network architecture for feature extraction,making the extracted image lighting feature maps more accurate.Secondly,a subspace attention denoising module is introduced after the second convolution of each convolutional chain,which achieves image denoising while weakening the optical effect of the image.(2)In order to solve the problems of excessive enhancement and severe loss of image details in the process of image enhancement,this paper proposes an improved dual fusion Unet curve estimation enhancement network suitable for nighttime traffic scenes.This network uses an improved Double Fusion Unet(DF-Unet)network as the base network for zero reference curve estimation enhancement,which integrates more feature information between the upper and lower layers during feature extraction,reducing the loss of detail information in the image to a certain extent.At the same time,in order to reduce the deviation of the enhancement effect and reduce the dependence of each enhancement operation on the previous enhanced image,this paper optimized the curve estimation algorithm by adding the original image as the deviation term of the algorithm in the enhancement operation.Through a large number of experiments,the weight of the Loss function is tested and adjusted.The improved network integrates more levels of detail information,which to some extent solves the problem of excessive image enhancement.(3)In addition,the article integrates the aforementioned suppression network and enhancement network,and designs an end-to-end image suppression enhancement algorithm based on curve estimation for night traffic image suppression enhancement.The input night traffic image is first processed through the SSA-De Glow network for image suppression and denoising,and then enters the DF-Unet network to extract the feature map of the image.Finally,the improved curve estimation enhancement algorithm is iterated to obtain the enhanced quality image.To verify the performance of the algorithm proposed in this paper,tests were conducted on several publicly available low light datasets,and the experimental results showed that the algorithm achieved good enhancement effects from both subjective and objective perspectives.In summary,this article comprehensively considers various factors that affect image quality and proposes a relatively complete algorithm for night traffic image suppression and enhancement.The various contents of the experiment indicate that this study has certain research significance for the development of computer vision technology.