The Research And Application Of LBM In Multiphysical Image Processing

Because of imaging mechanism of interferometric synthetic aperture sonar(InSAS) and ocean environment, the interferograms directly from InSAS alwaysinevitably contain a lot of noise. It will produce the results of subsequent phaseunwrapping. So, interferogram denoising is always an important subject in the studyof the InSAS system. In the interferogram processing, common denoising method isaverage filtering and median filtering. Recently, the partial differential equation (PDE)has been widely used in the field of image processing, and image denoising based onPDE is one of them. The image denoising method based on PDE which takesadvantage of the geometric feature of the image to lead the nonlinear diffusionprocess, can effectively remove the noise, at the same time, and also can keep theedge details very well. But, in order to guarantee the well-posedness of the model, itwill be pre-filtering inevitably. So, it can’t meet the requirement of the InSASreal-time processing. Recently, multiphysical image processing is just put forward. Ittakes advantage of multi-physics field to achieve the purpose of image processing.Firstly, This paper studies the image denoising method based on PDE, andperforms numerical experiments on the interferograms. Secondly, it Emphaticallystudied the lattice boltzmann model. Using the lattice boltzmann model, P-Mequation of regularization is evoluted and then performs numerical experiments on theinterferograms. Through the analysis of the result of the experiment contrast, wereached a conclusion: The lattice boltzmann model can effectively remove the noiseand also can keep the edge details very well. And, denoising efficiency of latticeboltzmann method denoising efficiency is much higher than P-M model ofregularization. Although, this method greatly improve the efficiency of denoising, itcan not exempt from pre-filtering. In view of this, this paper puts forward a model ofimproved Navier-Stokes equation, which is a multiphysical image processing model.Through numerical experiments on COMOSOL Multiphysics, We verify thefeasibility of this model on denoising. In order to further improve the operationefficiency of the model, this paper using the lattice boltzmann model evolve the new model. Through the comparison and analysis of the numerical results, we found thatthe lattice boltzmann model based on Navier-Stokes equations has much higherdenoising efficiency than P-M model, and it exempts from pre-filtering. Thus，itefficaciously meet the requirement of the InSAS real-time processing.The lattice boltzmann model based on improved Navier-Stokes equationseffectively improves the efficiency of the interferogram denoising, exempts frompre-filtering，and meet the requirement of the InSAS real-time processing very well.However, the model which needs to be in the next work to further improve alsoexists some problems in the process of numerical experiments.