Author(s): Anisia GOGU | Dorel AIORDACHIOAIE
Journal: Annals of Dunarea de Jos
ISSN 1221-454X
Volume: 32;
Issue: 1;
Start page: 48;
Date: 2009;
VIEW PDF
DOWNLOAD PDF 
Original page
Keywords: Signal Processing | denoising | filtering | thresholding | simulation | wavelets
ABSTRACT
The problem of image denoising based on wavelets is considered. The paper proposes an image denoising method by imposing a distortion input parameter instead of threshold. The method has two algorithms. The first one is running off line and it is applied to the prototype of the image class and it building a specific dependency, linear or nonlinear, between the final desired distortion and the necessary probability of the details coefficients. The next algorithm, is directly applying the denoising with a threshold computed from the previous step. The threshold is estimated by using the probability density function of the details coefficients and by imposing the probability of the coefficients which will be kept. The obtained results are at the same quality level with other well known methods.
Journal: Annals of Dunarea de Jos
ISSN 1221-454X
Volume: 32;
Issue: 1;
Start page: 48;
Date: 2009;
VIEW PDF
DOWNLOAD PDF Original pageKeywords: Signal Processing | denoising | filtering | thresholding | simulation | wavelets
ABSTRACT
The problem of image denoising based on wavelets is considered. The paper proposes an image denoising method by imposing a distortion input parameter instead of threshold. The method has two algorithms. The first one is running off line and it is applied to the prototype of the image class and it building a specific dependency, linear or nonlinear, between the final desired distortion and the necessary probability of the details coefficients. The next algorithm, is directly applying the denoising with a threshold computed from the previous step. The threshold is estimated by using the probability density function of the details coefficients and by imposing the probability of the coefficients which will be kept. The obtained results are at the same quality level with other well known methods.
