Spectral curvature destroys the co-registration of the spectra measured by dispersion imaging spectrometer.Using interpolation to re-sample the measured spectra at the non-offset mid-wavelengths can mitigate the spectral misregistration.It is very important to select an optimum interpolation method.The performances of six common interpolation methods are evaluated by comparing the residual errors in the corrected spectral radiance.The results indicate that, four-point cubic Lagrange interpolation and cubic spline interpolation are better than other four interpolation methods (linear Interpolation, three points quadratic polynomial interpolation, five points four-order Lagrange interpolation and cubic Hermite interpolation).For spectral offset of 10% Δλ(Δλ=5 nm), the normalized errors in measured spectral radiance is PV=0.06, that is reduced to PV<0.022 after interpolation with cubic Lagrange interpolation or cubic spline interpolation, but for other four methods they are PV>0.035.Furthermore, for lower spectral resolution (Δλ>5 nm), cubic Lagrange interpolation is a little better than cubic spline interpolation; while for higher spectral resolution (Δλ<5 nm), cubic spline interpolation is a little better.