Improved Image Denoising Research Based On The Median Filter And Wavelet Transform

Nowaday society has entered the era of digital information, while accounting for the largest amount of information storage is the image. Image signal becomes illegibility in the process of generation, transmission and recording due to disturbance from various noise, which will influence the later image segmentation, feature extraction, etc. Therefore we must preprocess the image to reduce noise before the following treatment, and image filtering is an important approach.There are many types of image noise, mainly represented as impulse noise and Gaussian noise- Median filter is widely used to remove impulse noise of a nonlinear denoising method; wavelet transform is an another powerful mathematical analysis tools after the Fourier transform, it has been widely applied in the field of image denoising and produced a lot of excellent wavelet image denoising algorithms based on the assumption of white Gaussian noise in recent years.In this paper, there are many in-depth studies of impulse noise, Gaussian noise filter which are common in the image- First of all, based on median filtering algorithm, combined with linear filters and nonlinear filters with their respective advantages, this paper made a linear-based, non-linear mixed-filter to improve the algorithm, called the hybrid median filtering algorithm, with the feasibility of experimental verification. Then did a certain amount of research work in the wavelet image denoising method to, the analysis of the characteristics of natural image wavelet coefficients on the basis of the given Gaussian mixture model based on wavelet image denoising algorithm with other wavelet denoising algorithm comparison experimental results show that the algorithm to remove Gaussian white noise have good results.Finally the actual images for common Gaussian pulse mixed noise, this paper will be mixed median filter and Gaussian mixture model based on the wavelet denoising algorithm combining image denoising algorithm. Simulation results show that being Gaussian, impulse noise pollution mixed images, the denoising effect of the algorithm is clearly much better than a single median filter or wavelet denoising method.