Research On Denoising Method Of Underwater Acoustic Image

As people continue to increase their exploration of the marine environment,images as one of the important ways to obtain information,underwater acoustic image processing has gradually become the research direction,This research plays a significant role in national defense military and civil engineering.Unlike optical images,sonar imaging is often affected by severe noise,so underwater acoustic images often exhibit low resolution,low contrast,and blurred target edges.How to effectively suppress image noise is the focus and difficulty of research,and it is also the basis for correct image analysis and understanding.This thesis focus on the features of low resolution and blurry edges of underwater acoustic images and how to improve the denoising performance of sonar images while maintaining the edge detail information.Based on the establishment of a variety of underwater acoustic images noise models,different denoising methods for underwater acoustic images are studied to form a method for obtaining high denoising ability and edge preservation ability for a wide range of environments.The main research contents are as follows:Establishment of sonar image noise model corresponding to ambient noise and reverberation noise.The formation mechanism and characteristics of the main noises affecting sonar imaging in different sea areas are discussed respectively.According to the environmental noise model and the scattering model of reverberation noise,a sonar image Gaussian noise model obeying the Gaussian distribution and a sonar image multiplicative noise model obeying the Rayleigh distribution are established.Research on transform domain denoising method.The denoising method of low rank decomposition in the transform domain is applied to the sonar image denoising in different sea areas.The Matlab experimental simulation shows that with the increase of the number of iterations,the denoising ability is enhanced while the target is fuzzy and cannot maintain the edge details of the target well.Research on spatial domain denoising method.In order to overcome the shortcomings of edge blurring after low-rank decomposition method denoising underwater acoustic image,Bilateral Filter,Guide Filter and Weighted Least Squares Filter are applied to sonar images denoising in different sea areas.The denoising effect of each filtering method under different noise variances in different sea areas is given by experimental simulation.The quality of image denoising of three filtering methods is compared by subjective evaluation and objective evaluation data.To conclude which method is more suitable for sonar image denoising,the denoising performance and edge-preserving ability of three filtering methods is compared.The results show that the de-noising ability and edge detail information retention ability of different sea,Guide Filter is better.Research on improved low rank decomposition method.Considering the good denoising performance and edge-preserving ability of the Guide Filter in the spatial domain,a sonar image denoising method combining the Guide Filter and low rank decomposition is proposed,and the parameters are analysed to optimize the performance of the method.By experimental simulation,the qualitative and quantitative comparison analysis among other denoising methods was carried out.The results show that the improved denoising method can overcome the shortcomings of low rank decomposition and preserve the target edge information while improving the denoising performance of the sonar image and the visual effect.