Research On Image Denoising Method Based On Wavelet Transform

With the development of the digital technology, digital image has been everywhere in our lives, and the image quality requirements of the public are also rising. But in the process of acquiring and transmitting of images, it always disturbed by many factors, and will inevitably produce some noises in this way. Therefore, image denoising becomes a quite important research content of the image processing.Image denoising is not a new topic, but people just have denoised images by some linear filtering method over a period of time. The proposeing of the wavelet theory, especially when the wavelet transform theory applied to two-dimensional image, the denoising problem of the image get a better solution, so as to promoting the development of image denoising. Denoising based on the wavelet transform means the noisy signal are transformed by the wavelet, wavelet coefficients of the getting signal and wavelet coefficients of the noise are displaying the different features in different scales, you can construct the corresponding rules to achieve the purposes of processing the wavelet coefficients of the signal and noise in the wavelet domain differently, reduce or completely remove those modulates produced by the noise, but retain the wavelet coefficients of the original signal to the utmost extent at the same time, finally, reconstruct the original signal by the inverse wavelet transform.The thesis is mainly based on the image denoising, introduce the basic theory of wavelet analysis, multi-resolution analysis and Mallat algorithm, as well as the traditional denoising method, principles and algorithms, and describe three image denoising method based on the wavelet transform:modulus maxima reconstruction denoising, the spatial correlation denoising and Wavelet threshold denoising, carry out a lot of simulation experiments on every methods, and thus obtain their respective advantages and disadvantages and application objects. Against on wavelet shareholding method, it has achieved advantages of relatively simple, small calculation and good treatment efficiency. Finally, analysis the wavelet shareholding method in more details, minutely discuss several crucial issues of selecting threshold and threshold function in particular. On this basis, propose an improved threshold denoising algorithm, and conduct simulation experiments on the platform of MATLAB. compared with the traditional denoising algorithm of soft threshold function and hard threshold function, conclude that it can eliminate image noise more effectively. And combine it with median filtering to construct new method, carry on numerical simulation to new method by using MATLAB, and conclude that the denoising have a better effect on the images of containing mixed noise.