Color characterization of display was studied by using polynomial regression and spectrum analysis. Once fitting and twice fitting between the RGB signals and the L*a*b* values were applied to establish color space conversion model. Through comparing the difference of ΔE and σ2, the differences of the models were analyzed. The results showed that the model can get satisfactory result when number of terms is 9 and 11; twice fitting is better than once fitting; when the term is 20, the model is not stable.