Image restoration by blind‐Wiener filter

Abstract

Wiener filter yields the minimum-mean-square error between the restored image and the original image. However, to obtain an optimal result, there must be accurate knowledge of the power spectra of the noise and the original image besides the degradation function. Otherwise, it will lead to an undesirable restored result. This study presents a so-called blind-Wiener filter that can restore the original image when we have no knowledge of the power spectra of both noise and original image. It uses the fact that averaging several consecutively measured images together will enhance signal-to-noise ratio (SNR). The number of images to be averaged to reduce noise to an acceptable level was concluded to be around ten. Ten independent random noises were added to a given corrupted image, resulting in ten images with different noises and then each of them was restored by Wiener filter to yield ten Wiener filtered images. Finally, the corrupted image was restored by taking an average over the ten Wiener filtered images. Experiments were conducted in a practical setting to demonstrate the effectiveness of the proposed method. The experimental results show that all the images in the test set were vastly improved and some images gave almost comparable performance to the traditional Wiener filter known as the best restoration method in terms of peak SNR.