We conducted a field experiment for two successive years with direct-seeded rice in the northern Sorachi region of Hokkaido, Japan, to understand the relationship between the normalized difference vegetation index (NDVI) measured by two types of optical sensors and the rice growth index from the fifth leaf stage to the heading stage, as well as to estimate the N uptake from the NDVI. A strong positive exponential correlation was observed between the NDVI and N uptake regardless of the type of sensor used. The correlation could be expressed by a single exponential formula regardless of the growth year or seeding patterns until the heading stage. The NDVI values obtained by the handheld crop sensor had a higher regression coefficient of the estimated N uptake (R2=0.88 and root mean square (RMSE)=15.4 kg ha−1) than those obtained by the satellite image (R2=0.68 and RMSE=24.9 kg ha−1). According to the regression curve between NDVI and N uptake, when the NDVI values were higher to a certain level and were saturated, N uptake considerably increased at a similar NDVI regardless of the type of sensor used. Therefore, the regression equations with the NDVI values obtained by the handheld crop sensor were recalculated when the R2 was maximum (NDVI<0.76), yielding an RMSE of 12.3 kg ha−1. Although the RMSE of the estimated N uptake by the NDVI obtained by the handheld crop sensor was slightly higher than the investigation of the growth index (plant length×tiller numbers×leaf color value) (R2=0.92, RMSE=12.7 kg ha−1), it could be a more practical method in terms of ease and investigation time.