Statistical Analysis And Denoising Method For Side Scan Sonar Image Noise

With the continuous development of sonar technology and the increasing demand for the exploitation of marine resources,it has become a hot issue to use side scan sonar to detect seafloor.In the process of detecting and imaging the seabed,in addition to the targets that can be detected,side scan sonar is also affected by sea-bottom reverberation,marine environmental noise,and some self-noise,in which the influence of reverberation is particularly obvious.Reverberation causes random distribution of spots on side scan sonar images,which is known as speckle noise.Speckle noise brings trouble to subsequent detection and recognition of side scan sonar images.Because of imaging principle of side scan sonar and complex environment of the seabed,side scan images show different structure characteristics and noise distribution with ordinary optical images.Therefore,it is of great significance to study the denoising algorithm for image characteristics of side scan sonar.In this paper,the statistical analysis and denoising methods of image noise of side scan sonar are studied,including the following three parts:Firstly,imaging principle and image characteristics of side scan sonar are analyzed,and considering reverberation statistic model of the side scan sonar,the fitting distribution of the Gamma distribution is optimized after using five typical probability distributions to fit the statistical characteristics of the seafloor reverberation.Then,estimation of optimal distribution parameters based on two feature parameters is proposed according to gray histogram.Through multivariate regression model,noise model of different sea-bottom types was established and the bottom classification was realized.On the basis of the above experimental results,the noise caused by different types of reverberation is simulated effectively by model parameters and image characteristics,and appropriate method is selected to verify noise simulation results.Finally,Field of Expert model is applied to image denoising of side scan sonar.A Field of Expert denoising algorithm based on Gamma distribution is proposed,which is called GammaFoE denoising algorithm.According to probability distribution model and Field of Expert model of sonar image,maximum posterior estimation can be computed by gradient rise method,and then image after denoising is solved.Among them,the parameters of the Gamma model are obtained by Expectation Maximization algorithm,and gradient-related method chooses the quasi-Newton method to obtain the fast convergence.Experiments show that GammaFoE denoising algorithm has good denoising effect on the side scan sonar images.