基于小波变换的正则化盲图像复原算法

Regularization algorithm for blind image restoration based on wavelet transform

提出了一种将小波变换和自适应正则化方法相结合的盲图像复原算法。该算法先对退化后的图像进行小波分解,得到图像在不同子频段的信息;然后针对各个子频段内图像的频率和方向特性,使用不同的自适应正则化复原方法,在图像的低频子频段进行去模糊;高频子频段则进行抑制噪声和保边缘特征;最后通过小波逆变换得到复原后的图像。实验结果表明,MSE减少了1.60,信噪比增量为1.76,算法性能和复原效果相对空间自适应正则化方法,都有一定的提高。

A wavelet based adaptive regularization scheme for blind image restoration is presented. The degraded image is decomposed to obtain its wavelet coefficients in wavelet domain, and the image＇s different frequency sub-bands are obtained also. Then, different adaptive regularization image restoration schemes are used in different sub-bands： removing blur in the low frequency sub-bands, while reducing noise and preserving edges in the high frequency sub-bands, and the algorithm finally obtains a restored image by adverse transforming. The experiments show that the MSE has a reduction of 1.60, while the SNR is increased by 1.76. It demonstrates that the blind image restoration method is more efficient compared with traditional space-adaptive regularization method.