The official testing methods for establishing nutritive values are accurate but relatively costly and time-consuming. Near infrared spectroscopy (NIRS) is potentially an alternative method that can analyze several components in a few minutes using an exclusively electronic instrument with no need for a laboratory expert. However, the accuracy of commercial NIRS spectroscopic food analyzers is not sufficient for Japanese food labeling, because of interference from moisture contained in the foods. This study aims to assess the effect of a freeze-drying pretreatment on the accuracy of NIRS food analysis. Thirty-four samples, consisting of six food items habitually consumed in Japan and cooked by different cooking methods were treated by milling then freeze-drying. They were analyzed by a commercial NIRS instrument (Calorie Answer^TM) with calibration curves developed based on other freeze-dried samples. The obtained nutritive values (energy, protein, lipid, carbohydrate and moisture) were corrected to the values before freeze-drying using the vaporized moisture content. The same samples before freeze-drying were also analyzed using the official testing methods to assess the analytical accuracy using NIRS after freeze-drying, and further analyzed using the same NIRS with the commercial calibration curves to assess the effect of freeze-drying. The accuracies were better for the freeze-dried samples than for the wet samples. The magnitude of the error in energy and carbohydrate was significantly associated with the retained moisture content in the freeze-dried sample. In conclusion, freeze-drying was an effective pretreatment for improving the accuracy of NIRS analyses of Japanese cooked foods, although it is still time-consuming and needs additional investment.