Research On Image Denoising Based On Median Filter And Wavelet Transform

Images are disturbed inevitably by all kinds of noise when collected and transmitted. In order to higher level processing, it's very necessary to carry on denoising pretreatment to images. There are many kinds of noise in images, mostly are impulse and Gaussian noise. Median filter is a nonlinear filtering technique widely used for removel of impulse noise. Wavelet transform is a strong mathematical analysis tool. It is widely applied to the field of image denoising in near years. And many good wavelet image denoising algorithms have appeared, on the assumption that noise of images is Gaussian white noise.This paper lucubrated the filtering of impulse noise and Gaussian noise, which are the most common in images. The research contents consist of three aspects:(1) We analyzed and studied standard median filter and its improvement algorithms. In order to overcome inherent shortcoming of standard median filter, the paper gives an adaptive median filter algorithm based on detection of impulse noise points, and compares it with other algorithms by simulation experiments. From the experimental results, we can know that compared with other median filter, the new algorithm can obtain the best effect when filtering impulse noise. (2) We did some research work about image denoising methods based on wavelet. After analyzing wavelet coefficient characteristic of natural images, a wavelet image denoising algorithm based on Gaussian Mixture Model is given. The experimental results indicate that the algorithm is better than other wavelet denoising algorithms when removing Gaussian white noise.(3) Usually, one denoising method is only effective to one kind of noise. In order to suppress mixed noise, the paper gives an image denoising algorithm, which combines the adaptive median filter based on detection of impulse noise points with the wavelet denoising based on Gaussian Mixture Model. Simulation experiments show that the algorithm is obviously much better than single median filter or wavelet denoising method when filtering mixed noise of Gaussian and impulse in images.