Research Of Wavelet Denoising Applications In Radar Image

With the development of science and technology, increasing number ofpeople pay more attention to the wavelet which is a new technique in the field ofnoise reduction.Radar image is not only used in the military field, now is alsoused in meteorology, people out for navigation, etc. The quality of radar imageinfluence people grasp of information accuracy. So in this thesis, the wavelet areanalyzed in radar image denoising.First of all, the thesis analyzes the image of radar, thereby discussing thenoise feature of radar image, the basic theory of the wavelet as well as thecharacter of the wavelet. MATLAB is a good software to simulate the radarimage with noises, the originally radar image can be added high frequency noiseswhich include the pepper noise and Gaussian white noise. Therefore it cansimulate the radar image which includes high frequency noises and originallyimage, thereby analyzing the relationship between wavelet transform and thenoise of radar images.According the information variable and the singularity of radar image, thispaper builds a new singular divisor p in the common threshold formula of thewavelet coefficient, meanwhile according equality between theory and reality. Athreshold value of radar images can be calculated. Firstly the radar image withthe pepper noise is decomposed one time by the wavelet, the threshold value cancut the high frequency areas, thereby achieve the function of noise reduction. Atlast the radar image without noise can be received after Wavelet restructuring. Atthe same time, this method can be compared with Median filtering denoisingmethod, which can verify the rationality and validity of the new thresholdformula.The method of Correlation wavelet denoising can aim at the Gaussian whitenoise which exists in the radar image. The radar can get a frame originally radar image during a period, each frame has the powerful correlation, especially twoframes which are the neighbor. According this character and the feature ofCorrelation between wavelet coefficients, the Correlation de noising algorithm isgenerated. Initially the experiment do the wavelet decomposition for the radarimage with the Gaussian white noise, then do wavelet threshold de noising forinformation which comes from different levels. The wavelet restructuring also isthe last step. Not only compare with the traditional Donoho threshold value, butalso do the comparison between the first level wavelet and the second levelwavelet, which can evaluate the de-noising effect and prove the rationality andvalidity of the wavelet correlation.