关键词: 图像复原, 小波变换, 多尺度分析, 正则化

Abstract: In order to restore blurred image to get the recognizable target，the adaptive regularized image restoration algorithm is adopted according to the detail character of target image.This paper emphasizes on discussing restoration method of the high degree blurred image.Through analyzing the limitation of some popular edge distribution estimation method which is used to restrain noise and faint edges，this paper proposes the method of image restoration based on wavelet multi-resolution analysis.Using wavelet-based multi-resolution analysis technique combined with morphology operator，the method works on the blurred image directly and obtains the edge weight image which can indicate edge distribution of original image reasonably.Last，the weight image is used as the adaptive condition to carry through image restoration.The experimental results indicate that the method can dispose the high degree blurred image effectively and help the restoration process converge quickly to gain the satisfying result.

Key words: image restoration, wavelet transformation, multi-resolution analysis, regularization