Study On Algorithm Of Image Denosing Based On Multiwavelet Transform

In the process of the image capture and image transmission, noise will be produced unavoidably. Thus, it becomes an important research project to denoise the noised image and enhance the image quality. For images denoising problems, the research on the WaveShrink method and its improved versions becomes one of hot research topics at present and it achieves some good results. The single wavelet basis do not has all the prosperities essential to the image denoising algorithm based on wavelet transform, which limits further improvement of the single wavelet WaveShrink algorithm in the image denoising effect. Hence, the multiwavelets which have better properties attract people's concern and some research results have been obtained. The denosing effects reported in the existing literature related are not ideal, due to not make full use of the characteristic properties of images in multiwavelets domain.In this dissertation we study the standard 256 level grey images, transform the images to multiwavelets domain by multiwavelets transformation, define a new adaptive WaveShrink function, using a special difference form of the Laplacian operator and and also considering of fractal dimension of every subband in the wavelet domain. And based on this, we present an Adaptive Multiwavelets Threshold (AMT) algorithm. Different from the former algorithms to denoising, the AMT algorithm can automatically determine the wavelet contraction threshold in the multiwavelets domain with none of empirical knowledge of images to denoise, the variance of the image noise, for instance.As for some standard 256 level gey images, series of simulated experiments are made. Experimental results show that, to the imbalancable multiwavelets, AMT algorithm not noly has good denoising effect, but also can effectively remain the detailed information veins characteristics of the images. Especially it has a remarkable effect for highly polluted images.