摘要
首先证明相干增强扩散算法(CED)与二维Haar小波收缩算法的等价性,推导图像像素点主曲率和主方向的表达式,并以之构造扩散矩阵。根据相干增强扩散与小波收缩的等价性,提出主方向各向异性扩散小波收缩算法。该算法在低频部分采用经典各向异性扩散(PM扩散),高频部分采用主方向各向异性扩散进行小波收缩。实验结果表明,主方向各向异性扩散小波收缩算法充分发挥了各向异性扩散算法良好的局部表现能力和小波收缩算法的快速计算能力,具有更好的降噪性能。
The equivalency of coherence enhancing diffusion （CED） and wavelet shrinkage are improved, the expression of principal curvature and principal direction are induced, and the diffusion matrix is construct by it. Based on the equivalency of CED and wavelet shrinkage, we put forward the principal direction anisotropic diffusion wavelet shrinkage image denoising algorithm. In the algorithm, the classic nonlinear diffusion method （PM） is adopted to process the low frequency part, diffuse the high frequency part using principal direction anisotropic diffusion to shrink the wavelet coefficient. The experimental results show that principal direction anisotropic diffusion wavelet shrinkage algorithm utilize the advantage of high-order anisotropic diffusion and wavelet shrinkage respectively, the denoising ability of it is better.
出处
《黑龙江八一农垦大学学报》
2008年第2期75-79,共5页
journal of heilongjiang bayi agricultural university
关键词
图像降噪
各向异性扩散
小波收缩
主方向
image denoising
anisotropic diffusion
wavelet shrinkage
principal direction
