Abstract: Segmenting canopy region effectively from the image captured by a digital camera is necessary for diagnosis of crop nutrition. An image segmentation thresholding method was proposed based on the difference of green and red channel (GMR) according to the difference of reflectance spectrum between vegetation and soil in visible band. The range of optimum threshold was from 10 to 20 during the rice growing stage, and the best threshold for shooting and booting stage was 10 and 20, respectively. The feature parameters of the segmentation of canopy images using this method are in good agreement with the SPAD readings, leaf nitrogen content, etc, in which the correlation coefficient between normalized redness intensity (NRI) and the two reached -0.87, -0.65, respectively. It demonstrates that this method is not only simple and accurate, but also applicable universally to other green crops.