Research On Image Enhancement Based On Retinex Method

Image enhancement is one of the most important aspects in image processing.Using artificial or machine to add some information to the original image,transforming some data features,highlighting some important features in the image,or suppressing some unnecessary features in the image,so that the image match human visual system.Enhancing the desired information in the image to make the visual effect of the image more in line with the human visual system.For a specific image,emphasizing the overall or some local characteristics of the image,the image will become clear if it is not clear.And improving the image quality,enhancing the readability of the image,to meet the needs of image analysis.For image enhancement,this paper does the following work:First of all,this paper introduces the image enhancement method : Retinex theory,and the process of its transformation into the field of mathematics.And one of the energy equations is changed to detect whether it can still enhance the image,which is model two.Putting forward the related minimum problem of energy equation,and proving the existence analysis of its solution.Next,the numerical experiments of the two models are given,that is,the finite difference scheme is solved by using the variational gradient downflow method of the model,and by using it,the numerical calculation is carried out.Secondly,giving the mathematical theory of the two models to analyze the existence of solutions.Introducing some theories and properties of function space,and use regularization method to prove the existence of variational solution.Finally,numerical experiments are carried out by using the existing finite difference scheme.Because of the experimental image is usually overexposed,a new parameter term is introduced based on the existing model parameters,which is convenient for the experiment.The experiment is divided into three parts.Firstly,the experimental results compare with Kimmel's experimental results and analyzed to prove that the two models are effective.Then,edge detection is carried out and compared with Mc Cann's method and improved Mc Cann's method to prove that the model can keep the original information of the image.Finally,the influence of the parameters on the experimental results is analyzed.