The Research On Image Denoising-the Application Research Of Image Denoising Based On 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.An image is often and inevitably corrupted by noise in its acquisition or transmission. The noise in an image has degraded severely the follow-up image processing tasks,such as image feature extraction, coding, segmentation, and target detection. Thus noise reduction becomes a very important image pre-processing for improving the quality of image and meeting the needs of higher level processing tasks. Median filter and wavelet filter in image denoising technique are mainly researched by this thesis.First,the first two chapters of the thesis play fundamental roles in the whole dissertation. The research background and significance of image denoising, and the development overview of image denoising, and the classification and mathematical model of image noise, and the evaluation methods of image filtering algorithm and the traditional image denoising methods are first introduced by this paper. Emphatically median filter algorithms and wavelet filter algorithms are introduced by this paper.Next, The conventional median filter, weighted median filter, switch median filter, maximum and minimum median filter and adaptive median filter are researched by this article. After their advantages and disadvantages analysised, an effective algorithm of image denoising for removal of impulse noise is proposed. According to the number of signal points in the noise detection window, the detection window size is changed adaptively by the proposed algorithm.Considered the correlativity of adjacent pixels, noise points and signal points are distinguished accurately by the proposed algorithm. And then according to the directional correlation-dependent, a edge-preserving filtering method is adopted by it to reconstruct the value of the corrupted pixel.The simulation results show that impulse noise in the image is removed effectively by the proposed algorithm, and the image edge details are well preserved by it, and better visual effects, and the proposed algorithm performs better than many of the prominent median filtering techniques reported.Then,wavelet modulus maxima filter method, wavelet coefficient correlation filter method and wavelet threshold filter method in the wavelet filter technique are researched by this article. After their advantages and disadvantages analysised, an improved image denoising method of first optimization and last classification in the wavelet domain is proposed. At first, the stein unbiased risk estimation used, an optimal threshold and the window of the neighborhood are selected by the proposed method for each sub-band in the wavelet domain,and then according to the size of neighborhood threshold,wavelet coefficients of sub-band are divided into "small" coefficients and "big" coefficients by it. Finally, those "small" coefficients are set to zero, while those "big" coefficients are modeled as zero-mean ganssian random variables with high local correlation, and the estimation of the true coefficients is obtained by minimum mean squared error criterion. The experimental results show that gaussian noise in the image is removed effectively by the proposed algorithm,and the texture information of image is effectively preserved, and better visual effects, and the proposed method performs better than many of the prominent wavelet filter techniques reported.Finally, the work is summaried seriously by this article and image denoising algorithms is made for further research prospects.