Multi-scales Image Denoising Method Based on Joint Confidence Probability of Coefficients

Abstract

Background: It is a classic problem that we estimate the original coefficient from the known coefficient disturbed with noise. Methods: This paper proposes an image denoising method which combines the dual-tree complex wavelet with good direction selection and translation invariance. Firstly, we determine the expression of probability density function through estimating the parameters by the variance and the fourth-order moment. Secondly, we propose two assumptions and calculate the joint confidence probability of original coefficient under the situation that the disturbed parental and present coefficients from neighborhood scale are known. Finally, we set the joint confidence probability as shrinkage function of coefficient for implementing the image denoising. Results: The simulation experiment results show that, compared to these traditional methods, this new method can reserve more detail information. Conclusion: Compared to the current methods, our novel algorithm can remove the most noise and reserve the detail texture in denoising results, which can make better visualization. In addition, our algorithm also shows advantage in PSNR.