基于非局部均值的彩色图像去模糊算法研究

Color image deconvolution based on non-local means algorithm

针对稀疏先验正则化方法非盲去模糊算法,在图像含噪声时去除模糊能力不强,提出了基于非局部均值算法与稀疏先验正则化相结合的非盲去模糊算法。首先对模糊图像进行高斯去模糊;然后使用域变换的边缘保存滤波,得到含有噪声的图像轮廓信息;再采用非局部均值算法去掉图像中的噪声信息;最后采用稀疏先验正则化滤波去卷积。实验证明,该算法能够有效的去除模糊,不会产生振铃效应,鲁棒性较好,具有较好的图像去模糊效果。

The image deblurring ability of non-blind deconvolution algorithm based on sparse adaptive priors is not strong when the image contains noise. A novel algorithm of non-blind deconvolution based on non-local means and sparse adaptive priors is proposed, in this paper. Firstly, deconvolve the blurred image by using Tikhonov regularization; Secondly, apply an edge-preserving smoothing filter with domain transform to obtain the image contour with noise; Thirdly, use non-local means algorithm to remove the noise; Finally,deconvolve the image by using sparse adaptive priors. The experiment results show that the algorithm has a superior performance without introducing border ringing artifacts. It＇s also very robust. It＇s an efficient approach to perform high-quality non-blind deconvolution.