Research On Image Optimization Algorithm Based On Prior Knowledge And NSST Transform

With the development of science,image information has penetrated into all aspects of life.It is a hot topic for many people to deal with images clearly and extract more feature information.Fog image and low illumination image belong to low quality image,which affects people's daily life and work to a certain extent.Therefore,it is necessary to study the restoration and enhancement of images.As a classical image restoration algorithm,the defogging algorithm based on the prior theory of dark primary color has important theoretical reference value for the research process of image.However,the algorithm has some shortcomings and needs to be further improved.The algorithm based on image enhancement is endless,and based on the classical algorithms related to contrast,brightness and hue,other mathematical models or theoretical framework can be integrated into one Step optimization results make the image closer to the reality and better visual effect.In this paper,two algorithms are proposed to restore and enhance the image in different directions.Finally,the effectiveness of the algorithm is verified.It is mainly composed of the following two parts:(1)Image restoration: Aiming at the deficiency of the prior defogging algorithm of dark primary color,based on the prior knowledge of foggy image,combining the method of large class variance,a new method of pre fog removal based on Ostu threshold is proposed.Based on the prior defogging algorithm of dark primary color,the four parts of image segmentation,atmospheric light value estimation,transmittance estimation and transmittance optimization are improved and optimized: for image segmentation: the threshold value is determined by using the improved double threshold segmentation method The image is divided into two parts:foreground and background by the threshold;for atmospheric light value estimation: according to the values of R,G,B color channels and the mental gray formula,the required atmospheric light value is obtained;for the transmittance estimation: Based on the different conditions after image segmentation,the reason of image distortion is analyzed and the best threshold is quoted The more accurate transmittance is estimated;for the optimization of the transmittance: the double index filter is used to optimize the transmittance,and the estimated value of the transmission is statically refined.The parameters are replaced by the frame of the prior defogging algorithm of dark primary color to obtain the defogging image.The results are compared with the four algorithms listed in this paper,which proves the feasibility of the algorithm.(2)Image enhancement: Aiming at the existing image enhancement algorithm in transform domain,based on Retinex theory,combining the non-Down sampling shear wave transform,an image enhancement algorithm based on adaptive threshold and local tone mapping is proposed.Based on the non-lower sampling shear wave decomposition of V component,the corresponding optimization methods are proposed for high pass and low pass subbands respectively: for high pass subbands of V component,a large number of noise is removed by using Bayesian shrinkage adaptive threshold method;for low pass subbands of V component,adaptive local tone mapping algorithm is used to compare them The degree of the V component is enhanced;the sub band of V component is inversely transformed by the non lower sampling shear wave,and the reconstructed V component is obtained and white balance is processed.Finally,the algorithm is verified by using low illumination image of flowers.The results show that the algorithm can optimize the image characteristics and the details have good recovery effect.In order to further verify the effectiveness of the algorithm,three test images are used to verify again.The results are compared with the three methods listed in this paper,which proves the feasibility of the algorithm.