Image Denoising Research Based On Nonsubsampled Contourlet Transform

With the rapid development of information technology,image processing technology has been applied in lots of field,which makes image processing theory and technology get rapid development.But when the hardware level is limited or the acquisition system is influenced by other factor,this will make the image that we obtain contain much noise,which reduces the image quality and vision effect;at the same time,this will disturb the subsequent research work,so image denoising is an important mission.The purpose of image denoising is to reduce the useless information and get the good denoising effect.This paper combines the multiscale geometry analysis with grey system theory,and conducts the image denoising research.The main research content of this paper:First,we introduce several multiscale geometry analysis,because the shift-invariant can eliminate the Pseudo-Gibbs phenomenon,so we introduce the nonsubsampled WBCT and nonsubsampled Contourlet transform;we select this two transform as the multiscale geometry analysis,at the same time we also introduce grey system theory and its some definition,which provides theory foundation for the subsequent research work.And then,we combine nonsubsampled WBCT with grey system theory,researching the correlation property of the three same direction coefficients in the same scale;according to the research of correlation property,our algorithm can recognize the noise point in the three same direction coefficients and denoise the noise;at the same time,we introduce the Total Variation theory to repair the low-pass image generated by nonsubsampled WBCT;and this has effective denoising effect.At last,we combine nonsubsampled Contourlet transform with grey system theory,researching the correlation property among the congenetic sub-band coefficients;according to the research of correlation property,our algorithm can select the strong edges,weak edges and noise in the sub-band coefficients and denoise the noise;at the same time,our algorithm can preserve the structure edge texture;we use the Total Variation theory to repair the low-pass image generated by nonsubsampled Contourlet transform,and then this can exert the function of the Total Variation theory very well;and this gets the effective denoising effect.