In two successive years (2003 and 2004), a set of 16 commercial sugar beet cultivars was established in Randomized Complete Block experiments at two sites in central Greece. Cultivar combination was different between years, but not between sites. Leaf sampling took place once during the growing season and leaf area, LA [cm2], leaf midvein length, L [cm] and maximum leaf width, W [cm] were determined using an image analysis system. Leaf parameters were mainly affected by cultivars. Leaf dimensions and their squares (L2, W2) did not provide an accurate model for LA predictions. Using L×W as an independent variable, a quadratic model (y = 0.003 x2 − 1.3027 x + 296.84, r2 = 0.970, p<0.001, n = 32) provided the most accurate estimation of LA. With compromises in accuracy, the linear relationship between L×W and LA (y = 0.5083 x + 31.928, r2 = 0.948, p<0.001, n = 32) could be used as a prediction model thanks to its simplicity.