Video Denoising Technology Based On Three-dimensional Wavelet Transform Of Surfacelet

The video which we daily see are all more or less mixed with some noise, how to wellremove the video noise has became a long time research focus of video processing, thepurpose of the video denoising in line with is to to improve the video quality, premise ofefforts to maximize retention of the useful signal and to maximize the removal of unwantednoise signal. Surfacelet transform is a real sense of three-dimensional wavelet transform, itcan effectively capture the singular information of curved which exists in themulti-dimensional signal, it is very suitable for image and video denoising. The denoisingresearch of Surfacelet transform isn’t depth as Contourlet transform which has the samestructure, the improved denoising methods which based on Surfacelet transformare not a lot,so it seems, Surfacelet transform has a lot of research space on denoising.In this paper, on the basis of Surfacelet transform the theory and practical applicationresearch, and focus on the Surfacelet transform video denoising. Through the research oftraditional Surfacelet transform video denoising by soft threshold function and hard thresholdfunction, from two directions to improve it, combined with other methods and proposed a newthreshold function, the concrete work is as follows.Due to Surfacelet transform is lack of translational invariance, its threshold denoisinggenerate pseudo-Gibbs phenomena and cause the distortion of reconstructed video image. Tosolve this problem, this paper applies Cycle Spinning to the Surfacelet transform videodenoising, many times pinning on the singal frame of the video image sequence, then throughthe reverse translation and average linear operation, the pseudo-Gibbs phenomenon is wellinhibited, while enhancing the PSNR of the video image after denoising.Although the PSNR results of traditional Surfacelet transform hard threshold functionvideo denoising is higher, but it is to a certain extent "over kill" the wavelet coefficients,making denoising image lost some details. A new threshold function is proposed on the basisof traditional soft and hard threshold function, the threshold function has large and small twothresholds, giving the secondary screening chance to the Wavelet coefficients which wereeliminated at the first time, to prevent the useful signal mistaken for noise,to prevent thephenomenon of the "over-kill". The application can be adjusted for coefficient thresholdingfunction according to the actual situation, get more suitable threshold function.