Research On Image Restoration Method Based On Texture Feature

In the process of image formation,shooting,and processing,imperfect imaging system,atmospheric environment,human operation and processing methods will cause loss of image details and decrease in resolution,which restricts the transmission and analysis of image information.Although some of the blur can be eliminated by improving the mechanical device and adding optical devices and other hardware.However,it is difficult to completely avoid the image quality degradation problems produced by some factors,whether in terms of hardware or environment,such as the diffraction limit of the optical system,the disturbance of atmospheric turbulence,and atmospheric scattering.Therefore,the restoration of degraded images from the perspective of digital image processing has irreplaceable value both in theoretical research and in practical applications.At present,many image restoration algorithms and schemes are proposed based on specific assumptions,and the actual blurred image may not be able to meet these prerequisites or only part of the conditions.Therefore,finding the characteristics and a priori that are more in line with the real image to guide the image restoration process is the key to solving the restoration problem and improving the restoration effect.The image texture is related to the brightness changes in the image,and can characterize the local and global structural characteristics of the image such as regularity,roughness and contrast.From the perspective of texture analysis,the research on the theory and key technology of blurred image restoration has important theoretical significance and application value.Based on the image texture characteristics,this paper improves the optical transfer function estimation model of the atmospheric turbulence degradation image,restricts the deconvolution process of the motion degradation image,and analyzes the content of the atmospheric scattering degradation image to obtain a more accurate estimate.First of all,this article summarizes the basic characteristics of image texture,respectively introduces the theoretical knowledge and applicability of texture description methods based on statistics,transformations,models and structures,and carries out the application of texture analysis in image processing and computer vision.A brief introduction is provided to pave the way for the algorithm proposed in this article.Secondly,the research on the restoration technology of atmospheric turbulence degradation based on the characteristics of texture frequency transform domain is carried out.This paper introduces the principle and characteristics of the classic point spread function estimation method,and improves the optical transfer function estimation model according to the frequency domain characteristics of the texture,and proposes a weighted estimation model incorporating the image gradient,and uses the signal energy of the image to restrict the restoration process.The algorithm improves the accuracy of the estimation of the optical transfer function,improves the visual effect of the restored image,and realizes automatic processing.Thirdly,the research of motion degradation restoration technology based on texture statistical features is carried out.This article summarizes the theoretical knowledge and mathematical forms of typical deconvolution algorithms,compares the restoration effects of each algorithm through experiments,and briefly analyzes their advantages and disadvantages.Aiming at the problem of motion degraded image restoration,the statistical characteristics of image texture are used to restrict the deconvolution process,suppress ringing artifacts,and obtain high-quality restored images.In this paper,a low-pass filter is used to generate a multi-scale image pair to detect and quantify ringing,and the quantization result is used as a regular term to constrain the generation of ringing at the edge of the image;then the number of iterations of each region is adaptively adjusted according to the statistical characteristics of the image texture.Further suppress the ringing effect.The algorithm can fully balance the contradiction between ringing suppression and edge enhancement in the deblurring process.Finally,the research on the restoration of atmospheric scattering degradation based on texture structure features is carried out.This paper discusses the key problems to be solved by the algorithm based on the atmospheric scattering model,analyzes the image content from the perspective of texture structure characteristics,constructs the image scene depth step map,and combines the quadtree search technology to propose an accurate estimation of atmospheric light The method prevents the estimated position of atmospheric light from being affected by white and flat objects and stably falls in a place with a large depth of field.In conjunction with the atmospheric light estimation method,this paper also proposes two solutions to correct the scene transmittance to solve the problem of atmospheric scattering degradation restoration.One is to perform scene analysis on the degraded image,and to estimate the transmittance of bright and dark scenes separately according to the centroid offset of the dark channel image to improve the accuracy of the estimation;the other is the global transmittance estimation method,which uses the expected value of the dark channel image to correct The transmittance of the whole scene,and the weighted residual map is established for the difference of the scene to enhance the restoration of image details.The accurate atmospheric light value and transmittance solve the problem of the overall brightness of the restored image being dark,hue shift and oversaturation.