Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

The Method of Image De-Noising Based on Wavelet Transform

Author ZhangQingWei
Tutor TaoChangLi
School Shandong University of Science and Technology
Course Basic mathematics
Keywords Image Denoising Wavelet Transform Radon Transform Ridgelet Transform Threshold
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 293
Quotes 0
Download Dissertation

Image is an important carrier of the information as well as an important channel of acceeding informations. However, the images are polluted by the noise or interferred by other non-target signals to different extents in every process of image acquisition, transmission, and access. In order to obtain the image informations more accurately, noise image need to be de-noised. Wavelet analysis is a new kind of frontier area. It has been attented extensively in signal and image de-noising while the wavelet analysis theory is improving daily. This paper mainly research on application of the theory of wavelet in image de-noising,the main contents is as follows:In the previous three chapters of this paper, we introduce the status of image de-noising,the basic theory of the waveletr analysis and the common image-denoising algorithms based on the wavelet transform. And we conclude the analysis and comparison about the three common methods of image de-noising based on wavelet transform.In the forth chapter,because the algorithm of image de-noising based on orthogonal wavelet transform should make the Gibbs phenomenon and the common threshold usually cause the tendency of over strangleding, we draw out the method of adaptive threshold image denoising based on stationary wavelet transform.We give out the adeptive threshold by correcting the common threshold based on different scale and sub-band direction because the signal and noise have different propagating characteristics. We show that this algorithm is reasonable and effective.In the fifth chapter,we introduce the ridgelet transform against the optimal basis of zero-dimensional singular objective function rather than the optimal basis of multi-dimensional objective function. Actually ridgelet is obtained by participating an orientation parpameter. The function of basis can describe the multi-dimensional singular signal along linear or hyperplane.We use ridgelet transform for image denoising because the linear singular of image is expressed by less ridgelet coefficients.But noise do not have so significant coefficients. So we can obtain better effect by proposing the method of adaptive threshold image de-noising based on ridgelet transform.We improve the common threshold according to the theory that the noise gradually weakened as the level of decomposition. Finally, we verify the effectiveness of this algorithm by experiments,especially to the image with features of linear singularities.

Related Dissertations
More Dissertations