Research And Application Of Image Denoising Method Based On Intelligent Algorithm

In digital imaging system,images often are degraded by noise because of the signal collection,transmission and imaging process,which will encounter various kinds of interference factors causing the degradation of image quality.During the process of advanced image processing,such as registration,segmentation and recognition,it always needs a distinct,high quality and clear target image.Therefore,image denoising has become an indispensable basic step in image processing field.Moreover,the efficiency of the image denosing directly affects the subsequent steps.The thesis is based on Spatial Numerical,Partial Differential Equations,and Multi-scale Geometric Analysis Method,which are the most important topics in image denoising.Although existing denoising methods can work well to a certain extent,denoising process often requires manual intervention in order to get better result which always involves adjusting the relevant parameters.Therefore,tharticle thesis introduces an Intelligent Algorithm(IA)to improve the adaptability of denoising process.as well as to get better result.The thesis mainly employs two kinds of the intelligent algorithm:Classification and Optimization.Classification based IA uses Extreme Learning Machine(ELM)as a classifier,and Optimization based IA using Genetic Algorithm(GA)and Particle Swarm Optimization(PSO)algorithm as the optimization tool.The thesis also studies how to integrate two approaches in order to achieve the purpose of image denoising.The major contributions of this work includes:1)the intelligent algorithm is employed to improve the algorithm performance;2)the intelligent algorithm integrates with the traditional denoising algorithm for image denoising and achieving ideal denoising effect.The major works are summarized as follows:(1)The denoising method on common salt and pepper noise in digital image,is studied via an adaptive denoising algorithm in conjunction with Extreme Learning Machine(ELM)algorithm as the classifier.The denoising algorithm uses ELM to differentiate pixels in the image,and then to remove the targeted noise points.The classification and handling of noise points can avoid a certain extent removing some image detail as image noise.At the same time,it can retain the original maximum limit in detail in the image in the process of denoising.The algorithm choose the ELM-RBF algorithm as a classifier,in the process of research found that although ELM-RBF has various advantages of ELM,but it also has some shortcomings.As the ELM-RBF learning algorithm random selects the center and width of the hidden layer nodes,in this paper,we use genetic algorithm to optimize the selection of the center and width of nodes,which gets a better classification effect and improves the efficiency of noise points detection.It is called GA-ELM-RBF algorithm.Then combining improved GA-ELM-RBF algorithm,ROLD factor and adaptive window size weighted average filtering method is applied to image denoising.The denoising simulation experiments prove the good denoising effect of method proposed.(2)The thesis also targets at another common Gaussian noise.,An adaptive denoising algorithm of partial differential equations is proposed by using Genetic Algorithm as the optimization tool.Firstly,the overall situation of Partial Differential Equation denoising method is studied,analyzes the denoising process of the Total Variational model and the Fourth Order Partial Differential Equation denoising process in detail,and discusses the advantages and disadvantages of these two methods.According to the characteristics of the two algorithms,this paper establishes the weight equation of two kinds of denoising results using the genetic algorithm to overall optimize the result of two kinds partial differential equation of the denoising method and presents GA-TV-4thPDE denoising algorithm in this paper which has a good denoising effect by denoising experiment.(3)On the basis of denoising research work of image Gaussian noise above,this paper proposes a denoising algorithm using Particle Swarm Optimization(PSO)algorithm as optimization tools.For better optimization results,first of all,the improvement of Particle Swarm Optimization(PSO)algorithm,this paper designs the improved PPSO algorithm with new particle rules of ' lowliest eliminate' and ' the same amount updating ',which combine with Performance Management(PM)thoughts.PPSO algorithm advances optimization speed and global search ability of the classical PSO algorithm.Integrating with the GA-TV-4thPDE algorithm proposed above,we use the improved PPSO algorithm instead of GA as optimization tool to remove the Gaussian noise in image.The denoising simulation experiment proved that this algorithm achieved good denoising effect.(4)Through the learning and analysis of Multi-scale Geometric Analysis(MGA)method of learning,this paper puts forward an adaptive image denoising algorithm removing gaussian noise of improved Shearlet transform algorithm.Choosing Shearlet transform as a basic denoising algorithm and combining with PPSO algorithm are to optimize shrinkage threshold selection of Shearlet transform denoising method.Firstly,through analyzing the advantages and disadvantages of Shearlet transform denoising method,it is found that the characteristic of multi-scale,more direction is good for expressing image structure and bring the difficult of choosing shrinkage coefficient.In view of this characteristic,we use the optimization strategy to obtain the optimal shrinkage threshold.Based on the research work above,this paper puts forward a denoising algorithm based on Shearlet transform,in the process of denoising.The PPSO optimization algorithm is used.Therefore,we got the PPSO-Shearlet algorithm in this paper.The denoising simulation experiment proves that the algorithm is effective and can achieve better denoising result.