Shanghai Jiaotong University, PR China.
Comput Med Imaging Graph. 2010 Mar;34(2):167-76. doi: 10.1016/j.compmedimag.2009.08.007. Epub 2009 Sep 25.
A novel denoising approach is proposed that is based on a spectral data substitution mechanism through using a mathematical model of one-dimensional singularity function analysis (1-D SFA). The method consists in dividing the complete spectral domain of the noisy signal into two subsets: the preserved set where the spectral data are kept unchanged, and the substitution set where the original spectral data having lower signal-to-noise ratio (SNR) are replaced by those reconstructed using the 1-D SFA model. The preserved set containing original spectral data is determined according to the SNR of the spectrum. The singular points and singularity degrees in the 1-D SFA model are obtained through calculating finite difference of the noisy signal. The theoretical formulation and experimental results demonstrated that the proposed method allows more efficient denoising while introducing less distortion, and presents significant improvement over conventional denoising methods.
提出了一种新的去噪方法，该方法基于通过使用一维奇异函数分析（1-D SFA）的数学模型进行谱数据替换的机制。该方法包括将噪声信号的完整谱域分为两个子集：保留集，其中保持谱数据不变；以及替换集，其中具有较低信噪比（SNR）的原始谱数据被使用 1-D SFA 模型重建的谱数据替换。根据谱的 SNR 确定包含原始谱数据的保留集。通过计算噪声信号的有限差分来获得 1-D SFA 模型中的奇异点和奇异度。理论公式和实验结果表明，与传统的去噪方法相比，所提出的方法在引入较小失真的情况下能够更有效地进行去噪，并具有显著的改进。
