Assuming that surface reflectance changes linearly with wavelength in the near-infrared band and the impact of path radiance is considered. An algorithm based on radiative transfer model MODTRAN, which gets the column water vapor amount using two lookup-tables (LUTs) is developed. Cubic interpolation method is used to reduce the water vapor amount step in LUTs. Using the LUTs through a stepwise-search method, water vapor amount distribution image can be gained from hyperspectral data. It shows that the radiance gotten by cubic spline interpolation only has 0.1% error compared with MODTRAN simulated radiance. Water retrieval is performed to two scenes of airborne visible/infrared imaging spectrometer (AVIRIS) data, and two water vapor amount distribution images are obtained. Reflectance data are obtained using the images through atmosphere correction. Water vapor amount distribution images change smoothly without bright or dark spots in them, and there is no peaks or valleys in the water vapor absorption band of reflectance spectra. It is proved that the algorithm is valid.