Research On Image Denoising Algorithm Based On Sparse Representation Theory

With the development of digital multimedia,people on image quality are increasingly high.However,noise pollution largely reduces the quality of the image,so image denoising is the focus of research in image processing.The traditional image denoising method is easy to realize,but it is easy to lose the texture and the structure information of the image.In recent years,with the popularization and development of compressed sensing,sparse representation theory has received much attention.In this paper,we mainly study the image denoising algorithm based on sparse representation theory,and improve the traditional sparse representation denoising algorithm,which has the following three aspects:First,the traditional image denoising method is easy to achieve,but easy to lose the image structure and texture information,according to the shortcomings of the traditional denoising methods,this paper puts forward a kind of based on Morphological Component Analysis decomposes image into structure and texture,the texture part use APDCBT dictionary sparse representation algorithm for denoising.The structure part use BM3D algorithm for denoising.Finally,the two parts are combined to get the final denoised image.Second,according to the traditional K-SVD image denoising method for signal utilization rate is insufficient,this paper uses Sparse Bayesian Learning of the image signal processing.Because the noise atoms exist in the dictionary after the dictionary is updated,the Bartlett method is used to cut out the noise atoms.Third,according to the traditional K-SVD method to run a long time and the sawtooth phenomenon of the steepest descent algorithm in convex optimization algorithmthe.In this paper,an improved steepest descent algorithm and orthogonal matching pursuit method are proposed,which can effectively solve the problem of running time and sawtooth phenomenon.In this paper,the simulation results show that the proposed method is better than the traditional denoising method to preserve the details and structural information of the image,and obtain higher peak signal to noise ratio(PSNR).Compared with the traditional K-SVD image denoising method,the proposed method can filter the noise and improve the running time.In order to get better results,the experimental results are compared with the Matlab GUI.