Research On Multiwavelet De-noising Method

Image, as an information carrier, has been utilized because of its visualization and information capability. However, images are often contaminated with noise during the process of generation or transmission due to a variety of effects. In order to get rid of the effect of noise and obtain the real information which the image is carrying, the noisy image needs to be de-noised. Whether a de-noising method is greatly affects the real and useful information attained from the noisy image. Although some existing methods were used to some extent, it was not ideal yet. This paper will illustrate image de-noising based on multiwavelet transform.Firstly wavelet transform, discrete wavelet transform and multiwavelet transform are introduced, then decomposition and reconstruction for image de-noising, thresholding shrinkage de-noising algorithm, de-noising algorithm based on translation invariant are analyzed based on the theory, and the wavelet transform modulus maxima de-noising algorithm and wavelet transform domain coefficients of correlation algorithms are introduced. By comparing the advantages and disadvantages of various wavelet transform and de-noising algorithms, translation invariant multiwavelet transform is used to transform noisy signal and image, and the thresholding function of thresholding shrinkage method has been improved, which is used in de-noising of noisy signal and image. The algorithm firstly carry on cycle shift of the noisy signals, in order to eliminate the dependence of wavelets in the time domain, and then the signal shifted is transformed by the multiwavelet transform instead of the traditional single wavelet transform, and thresholding processing is proposed on the transformed wavelet coefficients, using a new thresholding function which has benefits of both hard and soft thresholding function. Finally, the de-noised images are obtained through reconstruction of signal processed. Several existing classical algorithms and the proposed algorithm are compared in Matlab. Both experimental results and visual impression show the de-noising algorithm proposed in this paper is effective.