Assessment of grain quality or taste in rice is strongly required for regional production and marketing strategies of rice. For example, in rice production in several countries such as Japan, Korea, and Taiwan, the grain quality/taste is of higher priority than productivity these days. Remote sensing is promising for geo-spatial assessment of grain quality/taste over large paddy areas, which allows efficient and systematic soil/crop management as well as post-harvest processing/classification of products. We investigated the relationship between spectral reflectance and grain quality and growth of rice during ripening period. Hyperspectral reflectance data in rice were measured using ground-based and airborne sensors together with chemical contents of panicles and grain. Reflectance spectra were taken over 400-2500 nm wavelength region at 1 nm resolution in ground-based measurement (ASD-FR) and over 400-1050 nm region at 10 nm resolution in airborne measurement (CASI-3). We investigated the usefulness of whole- and selected-wavebands in partial least square regression method (PLS) as well as a range of conventional vegetation indices. Biomass, water content, and grain yield was well estimated using hyperspectral reflectance with PLS. Among the conventional vegetation indices the NDVI was one of useful indices for estimating protein content of rice grain at harvesting. Nevertheless, use of high resolution spectra provided mush better predictive ability in multiple plant and grain variables.