Sonar Image Denoising Based On Multi-band Wavelet Transform

Sonar system has been widely used in various kinds of underwater equipment, and it plays an important role in the military field. Sonar system is the most important way to access to target information. Sonar image reflect visual information of the target’s echo. There are all kinds of noise in the marine environment, they often mixed in the target echo and make the sonar image fuzzy. Then the noise can make the subsequent work be harder. So denoising is very important. There are many methods for optical image having a good effect. Imaging environment and imaging mechanism of sonar image is different to optical image’s. So the methods of optical image can’t be directly used in the sonar image. And we must find a method for sonar image. In the study of the past, gaussian white noise, a kind of additive noise, is the object of study. But speckle noise, a kind of multiplicative noise, is the primary noise of sonar image. In this paper, we take speckle noise as object of study and discuss it based on wavelet transform.The wavelet transform is one of denoising methods. And in recent years, it is widely used. The wavelet transform can distinguish noise wavelet coefficient with Mallat pyramidal decomposition based on multi-resolution analysis. The noise wavelet coefficients differ from the signal wavelet coefficients. The former is very much and very small, the latter is less and bigger. So according to this, we can reduce the noise coefficients to denoise the speckle noise, then at last we get the denoising image with wavelet reconstruction.The reason of widely application of the wavelet transform is that we can structure a variety of different wavelet function. Walsh matrix is a rapid transformation matrix composed of1and-1, having been used in image coding. Slant matrix is related to Walsh matrix, we can get it through Walsh matrix. We can take the different order Walsh matrix or slant matrix as coefficient matrix of different multi-band wavelet filter. In this paper, we firstly analyze the distribution characteristics of sub image based on multi-band wavelet transform of Walsh matrix or slant matrix. And then we set a series of different parameters, called inhibiting factor, to the different high frequency sub images. This method can minify the noise wavelet coefficient. So it can denoise the speckle noise. In this paper, we simulate the noising image with Matlab. And we perform a denoising experiment with the present method. Finally we get that the result of the present method is better than binary Haar wavelet.