摘要
文章提出了一种几何约束空间自适应阈值的图像去噪新方法。此方法基于非抽样小波变换的多分辨率分解,利用非抽样小波变换的冗余性来寻找小波系数之间的依赖关系。在BayesShrink阈值的基础上进行改进,采用空间自适应方法,为每一个小波系数确定自适应的阈值。在含噪系数的方差估计中,与以往的估计方法不同,不仅考虑到子带内小波系数之间的依赖关系,而且考虑了沿梯度方向的邻域内小波系数之间的依赖关系,使得含噪系数的方差估计更为准确。实验结果表明,与传统去噪方法相比,本文方法能更有效地去除噪声,具有更好的重建视觉效果。
A new image de-noising method which is based on geometrical constraints space adaptive threshold is proposed.This method is based on the multiresolution decomposition of nondecimated wavelet transform.The dependency relationship between wavelet coefficients are found by means of the redundancy of nondecimated wavelet transform.Improvement is made based on BayesShrink threshold and space adaptive method is used to give each wavelet coefficient an adaptive threshold.It is different from traditional estimate methods during the variance estimate of coefficients which contain noise.Not only the dependency relationship of wavelet coefficients within a scale is considered,but also the dependency relationship of wavelet coefficients within a neighborhood along the grads direction is considered.Therefore,the variance estimate of coefficients which contains noise is more correct.The result of experiment shows that compared with traditional de-noising methods,this method can eliminate noise more effectively and has better reconstruction visual effect.
出处
《科技信息》
2012年第11期116-117,共2页
Science & Technology Information
关键词
几何约束
图像去噪
非抽样小波
自适应阈值
Geometrical constraints
Image de-noising
Nondecimated wavelet
Adaptive threshold
