An adaptive threshold method for image denoising based on wavelet domain
Jiakun Xu, Kun Zhang, Mingyao Xu, Zhigang Zhou · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009
In this paper, a new thresholding function is proposed for image denoising in the wavelet domain. This function is used in an adaptive manner in a method that inspired form Thresholding Neural Network (TNN). Classic functions set the coefficients below the threshold value to zero, but in our proposed method these coefficients are tuned by a polynomial function. This tuning increases the capability of the function since we can attenuate the coefficients that are below the threshold value and close to it to a value less than the far coefficients. This function has some advantages over classical methods and produces better results in noise reduction. Besides the thresholding function, the subband-adaptive methods was adopted that the threshold value is selected differently for each detail subband. The simulation results show that the proposed thresholding function has superior performance compared to conventional methods when used with the proposed adaptive thresholding method. This makes it an efficient method in image denoising applications.