An Improved NMR Signal De-noising Algorithm Based on Wavelet Transform

The amplitude of nuclear magnetic resonance (NMR) longing signal general is very small while the signal to noise ratio (SNR) is also very low, so de-noising NMR signal before T2 spectrum inversion is necessary and important. In this paper, an improved de-noising algorithm based on wavelet transform is put forward. The main idea of this improved algorithm is that NMR signal is divided to several sub-window signals according to the exponential decay character of free induction decay (FID) signal and different sub-window signal is de-noised by wavelet transform using an improved threshold function. The detailed steps of this improved algorithm are discussed in this paper based on the steps of wavelet-based threshold de-noising method. In order to verify the performance of this improved algorithm, a 32-point small porosity model is designed and a free induction decay signal is generated by this model. Four algorithms are applied to filter the NMR noisy signal in an experiment: median filtering algorithm, wavelet transform filtering algorithm, FIR filtering algorithm and the improved algorithm, SNR and RMSE of these algorithms are also calculated in this paper. The result of this experiment displays that the improved de-noising algorithm based on wavelet transform, comparing to other algorithms, can improve the SNR of NMR signal and more effective keep useful signal.