There are several examples of relationships between NMR parameters and sensory characteristics of food products. Most of them are based on the measurements of the 1H NMR relaxation parameters T2 or T1 related to the texture or on the 23Na NMR relaxometry and 13C NMR spectrometry studies to explain the flavour. The present work describes an application of MR imaging techniques coupled with image analysis methods to predict the cheese texture characteristics given by a sensory panel. Six different types of cheese were selected. The cheese samples temperature was fixed at 15 °C ±1 for the sensory analysis and the measurements. 2D -FT MRI images of soft cheeses were obtained with a low field imaging system (Open Siemens, 0.2 T). Multi contrast images were acquired: Proton density T1 and T2 contrast images were measured with a spin echo sequence. Magnetic susceptibility images were measured with a gradient echo sequence. The MR image heterogeneity was quantified from the characteristic of the grey level distribution (mean, minimum, maximum, standard error, skewness, Kurtosis) and by texture analysis methods. From the numerous texture analysis methods, we chose the co-occurrence matrix algorithm proposed by Haralick. This method provides 13 parameters calculated from a statistical analysis of the grey level variations. A stepwise regression was performed to predict the sensory data from the texture parameters including the grey level distribution characteristics of all the MR images. The results obtained suggest significant links between the texture characteristic of the cheese and the properties of the MR images. From a linear combination of a limited number of texture parameters the most relevant sensory items were correctly predicted.