Image restoration via Wiener filtering with improved noise estimation
Hiroko Furuya, Shintaro Eda, Testuya Shimamura · International Conference on Signal Processing · 2009
In this paper, first, the performance of the Wiener filter in the frequency domain for image restoration is compared with that in the time domain for images degraded by white noise. After finding that the Wiener filter in the frequency domain is better than that in the time domain, a noise estimation method for the Wiener filter in the frequency domain is proposed. The frequency band division processing addressed recently is deployed and modified to improve the performance of the Wiener filter. The conventional noise estiation method does not estimate the noise spectrum in a low frequency region, and as a result it is not expected to restore a degraded image accurately. From this point of view, a method to estimate the noise spectrum in both low and high frequency regions is derived. The performance of the Wiener filter with the proposed noise estimation method is investigated through computer simulation experiments. Also, the optimal parameter values for the noise estimation method are searched. The proposed noise estimation method is compared with the conventional one based on Wiener filtering, resulting in the former providing a performance improvement.