Research On Image Denoising Based On Nonsubsampled Contourlet Transform And Partial Differential Equations

With the constant progress of science and technology, all kinds of image processingtechnology have been developing. Wavelet analysis has good time-frequency analysisand multi-resolution analysis capabilities, and is widely applied in the field ofimage processing. In recent year, as a mathematical theory which has a long history,Partial differential equations are applied in image processing. Images are oftencorrupted by diversified noises, image denoising have an important influence on thesubsequent image processing methods. The combination between wavelet analysisand partial differential equations are applied in image denoising has become aresearch focus. The traditional wavelet transform is shift sensitive and lacksdirection selectivity, nonsubsampled Contourlet transform that can better and moresparsely represent image, and it’s not shift invariance and more direction selectivity.Therefore, we study on image denoising based on nonsubsampled Contourlettransform and Partial differential equations. The main works are summarized asfollows:(1) We studies deeply on the relate theory of nonsubsampled Contourlet transform,including the Contourlet transform, nonsubsampled pyramid filter group andnonsubsampled directional filter group.(2) Researched several of model of anisotropic diffusion denoising based on partialdifferential equation, including the TV model and P-M model.(3) A novel image denoising method, based on nonsubsampled Contourlet transformand TV model, is given. This algorithm applies TV diffusion to low frequency, andthreshold shrinks the directional bandpass sub-band coefficients, experimentalresults show that our algorithm can make better ideal effect than any other methods.