Research On Blur And Impulse Noise Removal By Using The Framelet Based Image Decomposition Model

In the photoelectric imaging,image transmission,or storage process,the image quality is usually degraded due to the interference of the imaging environment and the imaging device itself.As a preprocessing process before image application,the image restoration process lays the foundation for subsequent image application analysis while improving image quality.Now it has become one of the primary problems in the field of image information processing.Image restoration refers to applying certain mathematical principles,combined with the image degradation process,to establish a corresponding image degradation and restoration model and then to solve the restoration model to obtain a clear image restoration algorithm.At present,image restoration includes research on deblurring,denoising,restoration,super-resolution,defogging,raining,and haze removal.This article mainly focuses on the image Gaussian noise,impulse noise,and image blur removal issues to do in-depth research.The thesis first studies the background and significance of image degradation and restoration and analyzes and summarizes the current research status of image deblurring and denoising at home and abroad.The paper examines the optical imaging process and the mathematical model of image degradation,introduces the compact wavelet frame,crossover algorithm,discrete cosine transform,split Bregman iterative algorithm,and quality evaluation indicators of restored images used in this research,and builds the paper The theoretical system.In response to the comprehensive problem of deblurring and impulse noise removal,researchers have successively proposed total variation(TV),non-local operators,low-rank and tight wavelet frame methods.Among them,the total variation and tight wavelet frame can protect the contour features of the image(the enhancement effect of the total variation is better).Still,the step effect of the total variation will lose the sharp edge features of the contour(such as distortion of the corners,sharp corners will Become smooth)instead of local operators,and low rank can protect the image texture information.Still,it is based on the principle of similar block matching.In image restoration,similar block matching errors will not be able to improve the image restoration quality further.In the research of deblurring and denoising the image simultaneously,the existing methods are to study all the information of the image as a whole,which will make the restored image lose a lot of important information.However,according to recent image analysis theories,an image contains multiple layers of information at different scales:contour features at a large scale and texture information at a small scale.A model can only effectively characterize one layer of information and will lose its utility when characterizing other layers of information.Therefore,to restore the contour and texture information of the image at the same time,strengthen and protect the edge features of the image contour,and avoid the step effect,the paper proposes a simultaneous deblurring and de-impulse noise model based on the multi-layer decomposition of the image,which achieves a high level of the restored image—quality improvement.The paper uses a tight wavelet framework to characterize image contour information,and discrete cosine transforms to characterize texture information to achieve effective image decomposition.In order to enhance contour edge features and avoid step effects,the paper adds a total anisotropic change to the regular term of the contour.The paper uses a crossover algorithm and split Bregman iteration to solve the proposed multivariate image restoration model and derives the corresponding algorithm.To prove the effectiveness of the proposed algorithm,experimental research is carried out,as well as comparative experiments with other methods.The experimental results prove that the model and algorithm proposed in this paper can effectively remove the blur effect and impulse noise of the image simultaneously and can further restore the contour layer and texture layer information of the image at the same time—the paper designs the corresponding image processing system based on the proposed model and algorithm.