The Application Of L0 Norm In Image Enhancement Variational Model

As an effective way for humans to perceive and understand the world,images contains many visual information.However,due to the fact that images are acquired and transmitted through many processing steps.they are subject to many external factors.This not only affects the visual effect,but also affects the subsequent processing.Therefore,the enhancement of the image is particularly important.Recently,The image enhancement technology is becoming more and more mature,and image enhancement techniques based on variational are becoming more and more popular.Especially after the introduction of the Retinex variational model,many scholars have proposed improved models,and the TV-Retinex model has been widely applied to image enhancement.Similarly,the use of atmospheric scattering models for image enhancement is becoming more widely used,especially in image defogging.In this paper,we study these two classic image enhancement methods in detail,and use the sparsity of L0 norm to improve the classic image enhancement algorithm as follows:Firstly,based on the TV-Retinex variational model,this paper proposes an image enhancement variational model based on L0 norm and Retinex theory.Since the L0 norm has better sparsity than the L1 norm,the regular term constraint of the illuminance component is improved from the L1 norm to the L0 norm and applied to the Retinex variational model.The experimental results show that the illuminance component estimated by the proposed algorithm is smoother.At the same time,the estimated reflection component can distinguish more structural information while retaining more details.Secondly,an image enhancement variational model based on L0 norm and atmospheric scattering model is proposed.The traditional image dehazing variational model uses the L1 norm as the constraint of the gradient regular term.We use the sparsity of the L0 norm to the regular term.For the solving problem of L0 norm,the L0 gradient minimization algorithm is proposed and the detailed solving steps are given.The improved model proposed can preserve the structural texture information of the image while removing the haze.