During the drying process, internal water migration is often the limiting factor for the overall water transfer. In order to simulate a drying process, an accurate determination of the water diffusivity (D) in the material is always necessary. For biological products D is known to be low and vary with moisture content (X) and temperature (T). D = f(X, T) can be estimated from the fitting of moisture profiles determined by non-intrusive NMR imaging. This work first presents the experimental results obtained on paddy rice. The moisture profiles were measured during the drying process using a Bruker AMX400 spectrometer, equipped with a micro-imaging device at the H+ frequency of 400 MHz. Four drying experiments at two different air temperatures were performed. A constant time imaging (CTI) technique proved to be useful to obtain the moisture content map in a central slice section of the kernel with 3 turn thickness, allowing a spatial resolution of 0.1 mm. This method provided access to low moisture content and low water mobility data. Then a diffusive model was developed using a cylindrical geometry, taking into consideration the shrinkage during drying. Uni-dimensional water profiles from axes of the elliptic section of the kernel were selected in order to determine the diffusivity parameters of the drying model.