Power spectrum estimation method for image denoising by frequency domain Wiener filter

In this paper, a new denoising method for images degraded by white noise in blind condition using the Wiener filter has been proposed. The Wiener filter is a well-known denoising technique, but the performance comparison of the Wiener filter in different domains is not clear. This paper investigates and evaluates the Wiener filter implemented in the frequency, time and wavelet domains in an ideal case. It is shown that the frequency domain Wiener filter outperformes the Wiener filter implemented in the other domains. In addition, this paper proposes a power spectrum estimation method to be employed with the frequency domain Wiener filter. The performance is studied and found that the proposed method is capable to denoise noisy images in a wide range of noise levels and provides improved or equivalent performance relative to conventional methods.