摘要
尽管多小波相比于单小波,具有正交性、紧支撑、实对称、高阶消失矩等性质,但是多小波进行图像去噪效果并不是很理想,其主要原因是没有充分利用图像在多小波域内所特有的性质.通过将含噪图像变换到多小波域,在小波域内应用Laplace算子的一种特殊差分格式,并考虑小波域内各个子带的分形维数,提出了一种自适应的多小波阈值算法,即AMT算法.对于欲去噪的图像,AMT算法能够在多小波域内自动确定去噪的小波伸缩阈值,而不需要知道图像的任何先验知识,例如图像中噪声的方差等等.仿真实验结果表明,该算法去噪效果好,特别是对于高度污染的图像,AMT算法的去噪效果更加显著.
Multiwavelets, in comparison with single wavelets, have properties such as orthogonality, short support, real symmetry, high order vanishing moments, etc. However, they do not do well in image denoising, primarily due to algorithms that fail to make full use of the characteristics of images in the multiwavelets domain. This paper proposes an adaptive multiwavelets threshold algorithm, or AMT algorithm. This transforms a noisy image into the multiwavelets domain, applies a special difference format of the Laplace operator, then considers the fractal dimension of every sub-band in the multiwavelet domain. Before a noisy image is denoised, the AMT algorithm automatically determines the wavelet threshold in the multiwavelets domain without empirical knowledge of the image, for instance, variance of the image noise. The result of simulations reveal that the denoising effect of the AMT algorithm is measurably better, especially for very noisy images.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2007年第5期594-598,共5页
Journal of Harbin Engineering University
