The Research Of Underwater Image Denoising Based On Wavelets

It is well known that the 21st century is the age for oceans'development.For China with large population,limited land sources and worse and worse environment,to a extent,abundant oceans sources can largely alleviate various predicaments confronted by China at present,so it is a project with important significance and profound impact to develop seas.However,nowadays people understand oceans very little.Therefore,how to exploit the treasure of unlimited potentiality reasonably and effectively is a difficulty problem placed in front of scientists.Because of complicated underwater environment and bad imaging contitions,images from underwater tend to be flooded in all kinds of noise.The paper's goal is to try to use wavelet,the rising tool ,to denoise for noise picked up from true underwater images,in order to raise images'contrast ratio and improve images'quality.Firstly,the paper studies the fundamental theory of wavelet transform ,analyses wavelet transform's characteristics and later expounds and compares three main methods for denoising in wavelet domain.The paper emphasises threshold denoising methods in wavelet domain,among which several classical ones that estimate thresholds are indicated in detail. Secondly,the paper analyses briefly the property of underwater images'background noise.On the basis of this,according to the idea of"level-independent thresholding",classical methods for estimating thresholds under all kinds of standards are combined and images with underwater noise are denoised.The paper proposes a new method for estimating noise variance to determine thresholds.Denoising effects of two methods are compared and the reason is discussed.Finally,the method of fixing thresholds based on generalized cross validation standard is studied.As well as built on the basis of the idea of"level-independent thresholding",the method is used to denoise images with underwater noise.Because orthogonal wavelet transform has the shortcoming of lack of linear phase and translation invariant,generalized cross validation with redundant wavelet transform and biorthogonal redundant wavelet transform are respectively combined to denoise.Experimental results show the denoising effects of the two latter methods are better than the one of the former method.