The hyperspectral reflectance of soil was measured by a ASD FieldSpec within 400~1 000 nm.Next, its first derivative of spectra were acquired and de-noised by the threshold de-noising method based on wavelet transform. From the de-noised derivative spectra, absorption areas used as indicatoresas for soil organic matter content were acquired by numerical integration. Results show that: (1) Because of much noise, it is difficult to identify spectra contour and features in the first derivative of soil spectra resulting from different organic content levels. (2) When the scale of wavelet decomposition was 3, the threshold de-noising method based on wavelet transform can keep the balance between smoothing curve and holding spectra features. (3) Absorption area S(538, 586) is extracted from de-noised first derivative of soil spectra, and the coefficient of correlation between it and organic matter content is 0.896 3.