Novel denoising technique based on filtering the coefficients of a redundant wavelet transform

A denoising technique based on filtering the coefficients of a redundant wavelet transform is proposed for the purpose of reducing statistical noise in images without blurring sharp edges. The wavelet coefficients in the higher-spatial-frequency components are denoised using a function that has been used in nonlinear diffusion filtering. With this function, edge and noise points were correctly distinguished. This resulted in the elimination of the wavelet coefficients at the noise points and the preservation of the wavelet coefficients at the edge points. As a result, most of the image noise was successfully eliminated without appreciable losses in the sharp edges. Thus, the denoising technique coped with both noise reduction and edge preservation.