Research On Enhancement Methods Of Uneven Illuminating Images

Due to the limitation of imaging conditions,the images taken in real life often have uneven illumination.It leads to low contrast,color distortion and detail loss of the image,which not only affects the visual effect,but also makes some computer vision algorithms(such as object recognition,object tracking,etc.)difficult to work.Based on Retinex model,an adaptive dark region detail enhancement algorithm was designed.Retinex theory decomposes the image into a illumination map representing the scene illumination and a reflection representing the physical characteristics of the scene.In this paper,the formula of dark image enhancement based on Retinex is firstly transformed into a bayesian optimization framework.Then,the dark region prior and adaptive gradient amplification strategies are designed,and the prior is integrated into the estimation of the illumination map.The regularization of smoothness of dark region and structure perception is optimized by using the dark region prior and the widely used spatial smoothness and structure prior.In this paper,an algorithm is given to solve the optimization problem,and then the enhanced results are obtained through post-processing.Experiments show that the algorithm can get better enhancement effect and show more dark area details.An enhancement algorithm for fusion of visible and near-infrared images is designed.The algorithm is designed to take advantage of the details of the scene typically contained in near-infrared images,which are difficult to detect in visible images.In the illumination estimation process of the Retinex model,the gradient information of the near-infrared image is incorporated into the structure prior,combined with the spatial smoothness prior to obtain the optimized illumination map,and then the final enhancement result is obtained after post-processing.Experiments show that the fusion of near infrared image information can restore the visible light image well.