Selecting an appropriate sample size is quite important to achieve acceptable accuracy in evaluating vast natural areas with field and remote sensing methods. This study aimed to evaluate the effect of different sample sizes on the correlation of vegetation cover and remote sensing-based indices such as NDVI in three rangeland types, Semirom, Isfahan. Landsat5 (TM sensor) and IRS-P6 (AWiFS sensor) satellite images obtained at 1/6/2009 and 22/5/2009 ,respectively, were used to conduct this research. The percentage of vegetation and phisical ground components including canopy cover, litter, rock, gravel, stone and bare soil was determined using step-point method in 4 radial direction in 30, 50, 100 and 150 meters (1200, 2000, 4000 and 6000 points per rangeland type). In other words, the ground cover and NDVI indices correlations were compared in 2826, 7850, 31400 and 70650 square meters using the satellite images. NDVI indices were calculated in all vegetation types and  correlated with different components of ground cover. According to results, TM sensor data had more appropriate correlation than AWiFS sensor with field vegetation cover in all studied sample sizes. Vegetation cover and NDVI correlations were decreased in degraded vegetation types with increasingground heterogeneity. For example,  the correlation was not significant (a= 0.05) in all studied sample sizes in degraded Astragalus spp – Scariola orientalis vegetation type, while the highest correlation was obtained (77 percent) in 7850 square meters sample size in more homogeneous Bromus tomentellus vegetation type.