基于尺度空间相关的自适应小波滤波算法研究
Adaptive wavelet threshold for signal denoising based on the scale space correlation
祝传广 1邓喀中 1范洪冬1
作者信息
- 1. 中国矿业大学环境与测绘学院,江苏徐州221116;江苏省资源环境信息工程重点实验室,江苏徐州 221116
- 折叠
摘要
针对信号滤波中噪声方差的估计、阈值的确定及细节信息的保留三个问题,利用添加高斯白噪声的模拟信号进行实验分析,结果表明:噪声主要集中在低尺度,因此可以在低尺度空间估计噪声方差；噪声的能量随尺度的增大而逐渐减弱,阈值采用自适应算法确定较为合适；小波自适应滤波残留较多的小噪声,因此提出了尺度相关与小波自适应相结合的滤波算法.该算法的滤波结果表明,信号的信噪比较之传统小波算法可以提高10％.并且由目视效果可以看出,通过该算法滤波后的信号较之传统小波算法更加平滑.
Abstract
The simulate data was used to analyze three problems in the signal filtering, the variance of the noise, the threshold and the extracting of detail features. And the result shows that a) the wavelet coefficients of the noise congregate in the low scale, and the noise variance can be calculated in the finest scale therefore, b) the power of noise reduce with the scale increasing, and thereby the threshold should be different in different scale, c) although the adaptive wavelet threshold can filter most of noise, much noise still remain in the filtered coefficients in order to protect the detail informa-tion. The signal and noise had different statistical characteristics, and based on which the scale space correlation filtering can extract signal from the noisy signal. But if there were much noise, the scale space correlation filtering method performed bad because many of coefficients of signal was silenced by noises. So combined filtering method was combined the scale space correlation with the adaptive wavelet threshold to filter the signal. And the experimental results show that the combined filtering method can increase the SNR by 10% comparing to adaptive wavelet method. Besides, the filtered signal is smoother than the adaptive wavelet method.
关键词
滤波/小波/尺度相关/自适应/阈值/信噪比Key words
filtering/wavelet/scale space correlation/adaptive/threshold/signal-noise ratio引用本文复制引用
基金项目
国家自然科学基金(41071273)
中央高校基本科研业务专项基金(2010QNA21)
高等学校博士学科点专项科研基金(20090095110002)
国土环境与灾害监测国家测绘局重点实验室开放基金(LEDM2011B07)
出版年
2012
NETL
NSTL
维普
万方数据