摘要
非局部总变差本质上是一种局部去噪方法,该方法的不足是无法利用整幅图像的冗余性来进一步减少噪声方差。为获得一个更加真实有效的非局部解决方法,以傅里叶变换为依据,在非局部总变差的基础上提出一种新的变分模型,这种方法称为空间频率域非局部总变差。该算法归结为最小二乘数据拟合、空间频率域非局部总变差和小波系数正则化的最小化线性组合。在模型求解方面,利用快速组合算法可实现快速收敛,提高算法的求解速度。经仿真验证,该算法处理速度更快,图像复原效果更好,可用于压缩磁共振图像复原中。
NLTV is essentially a local denoising scheme and the shortcomings of this approach is unable to use redundancies in the whole image to further decrease the noise variance.In order to obtain a more realistic and effective nonlocal solution,this paper propose a new variational model by adding to NLTV a term based on the Fourier transform,we call this new scheme spatial-frequency domain nonlocal total variation（SFNLTV）.This algorithm formulated as the minimization of a linear combination of three terms corresponding to a least square data fitting,spatial-frequency domain nonlocal total variation（NLTV） and wavelet sparsity regularization.In term of solving the model,using fast composite splitting algorithm（FCSA） can achieve fast convergence,improve the speed of solving the algorithm.The simulation show that the processing speed of the algorithm is fast,and obtained better reconstruction result,can be applied to compressed magnetic resonance（MR） image.
作者
张宝强
蔡述庭
Zhang Baoqiang Cai Shuting(Institute of Automation, Guangdong University of Technolog)
出处
《自动化与信息工程》
2016年第4期35-39,48,共6页
Automation & Information Engineering
关键词
磁共振图像
非局部总变差
傅里叶变换
快速组合分裂算法
Magnetic Resonance Image
Nonlocal Total Variation
Fourier Transform
Fast Composite Splitting Algorithm
