摘要
传统单幅图像去模糊方法需要稀疏先验约束,导致计算量较大。为此,在自适应最小均方误差(LMS)算法的基础上,提出一种点扩散函数(PSF)估计方法。利用模糊图像得到有效突出边缘,作为自适应滤波器的输入信号,并将模糊图像作为滤波器的期望信号,用以估计PSF。在非盲去卷积过程中,采用各项异性正规化方法对清晰图像进行约束,以减少恢复图像的振铃效应。实验结果表明,该方法不需要先验约束,对运动和非运动模糊图像均可适用,在保留图像细节的同时能抑制平滑区域的噪声。
The traditional method to deblur single blurred image required a variety of sparse priori constraints, in order to solve this problem, an adaptive Least Mean Squarc（LMS） error algorithm for getting the Point Spread Function（PSF） is proposed. This algorithm does not require priori constraints, in the case of only a blurred image, First, get an effective strong edge of the latent image as the input signal of the adaptive filter, while blurred image as the desired signal, then estimate the PSF; In the non-blind deconvolution process, in order to reduce ring artifact of the restored image, an anisotropic regularization constraint term on the latent image is adopted. The experimental results show that the PSF estimation method not only applies to motion blur image, but also applies to defocus blur image and uniform blur image.
出处
《计算机工程》
CAS
CSCD
2012年第17期226-231,共6页
Computer Engineering
基金
国家自然科学基金资助项目(61062014)
广西自然科学基金资助项目(桂科自0991280)
关键词
双边滤波
冲击滤波
自适应LMS滤波
点扩散函数估计
图像恢复
最大似然估计
各项异性正规化
bilateral filtering
shock filtering
adaptive Least Mean Square（LMS） filtering
Point Spread Function（PSF） estimation
imagerestoration
maximum likelihood estimation
anisotropic regularization
