The　principles　of　three　types　of　wavelet-based　methods　are　discussed　and　compared　in　low-field　pulsed　NMR　signal　denoising.　The　wavelet　threshold　denoising　method　based　on　Stationary　Wavelet　Transform　(SWT)　is　relative　brief,　which　can　restrain　high-frequency　noise　greatly　while　leave　some　lowfrequency　noise　unsettled.　The　denoising　algorithm　based　on　modulus　maximum　in　wavelet　field　can　suppress　both　high-　and　low-frequency　noise,　while　some　peaks　are　weakened　in　the　denoising　process.　An　improvement　is　presented　to　spatial　correlation　denoising　algorithm　by　translating　and　lining　up　the　edges　ahead　of　multiplying　the　adjacent　wavelet　scales,　which　is　proved　in　our　experiments　more　effective　to　remove　both　high-and　low-frequency　noise　greatly　while　protecting　peaks　well.　©　2014　WIT　Press.