Research And Application Of Image De-noising Algorithm Based On Wavelet Transform

The research of image de-noising algorithm based on wavelet transform is one of the hot topics in the image processing field both at home and abroad. This paper studies the wavelet thresholding algorithm, and verifies the validity and feasibility of the algorithm by processing the medical image and SAR image. The main contents include the following:(1) MLE algorithm for Gaussian noise of the image is provided to estimate the noise level of the image. Firstly, according to the characteristics of the Gaussian noise model, the author applies the Maximum Likelihood Method to estimate and analyze the noise model contained in the image. Secondly, the noise images are represented with histograms and different samples in the normalized histogram are selected. Then the noise variances are estimated by using the MLE. Finally, the simulation experiment is carried out in the Matlab simulation environment. Experimental result shows that the variance of the actual image and the noise variance obtained by this method are approximately equal. Therefore, the algorithm has excellent features both in feasibility and accuracy.(2) An improved threshold function is proposed for the problems such as distorted de-noised image, the appearance of Gibbs concussion and so on, which are caused by the discontinuity at the threshold and the deviation of wavelet coefficients. Compared with conventional hard threshold, soft threshold and the existing threshold function, this function is not only easy to calculate, but also has excellent mathematical properties. To verify the superiority of the threshold function, the author compares the PSNR and MSE produced by existing de-noising methods with this algorithm. Simulation experiment result shows that this de-noising method is superior to common threshold function in visual effects, analysis of variance and performance of PSNR and MSE.(3) In order to improve the demerits of the wavelet thresholding function and drawbacks of de-noising by using a single threshold value, the paper puts forward new hierarchical threshold and threshold function. Firstly, the noise image is wavelet decomposed to obtain wavelet coefficients. Secondly, the high frequency coefficient is processed by the improved method. Finally, based on the obtained wavelet coefficients, the images are reconstructed under the condition of wavelet basis function to get the de-noised images. The medical image simulation experiment shows that the de-noising effect of this algorithm is superior to the common threshold function in visual effects, analysis of variance and performance of PSNR and MSE. Therefore, this method should be promoted and applied in solving practical de-noising problems. An adaptive wavelet thresholding algorithm is proposed to the problems of image distortion and edge blur, which existed in the de-noised SAR image of traditional thresholding methods. Firstly, the logarithmic transformed SAR images are wavelet transformed. Secondly, the high-frequency of the conversed SAR image is de-noised by using adaptive wavelet threshold. Finally, the inverse wavelet transform and the exponential transform are processed. Experimental result shows that the algorithm can not only get better de-noising effect but also avoid the shortcomings of traditional algorithms effectively.