The amount and distribution of gaps in vegetation canopy is a useful indicator of multiple ecosystem processes and functions. We describe a semi-automated approach for estimating canopy-gap size distributions in rangelands from high-resolution (HR) digital images using image interpretation by observers and statistical image classification techniques. We considered two different classification methods (maximum-likelihood classification and logistic regression) and both pixel-based and objectbased approaches to estimate canopy-gap size distributions from 2-3cm resolution color infrared aerial photographs for arid and semi-arid shrub sites in Idaho, Nevada, and New Mexico. We compare our image-based estimates to field-based measurements for the study sites. We found a strong relationship (R2 > 0.9 for all four methods evaluated) between image- and field-based estimates of the total percent of the plot in canopy gaps greater than 50cm for plots with a classification kappa of greater than 0.5. Performance of the four remote sensing techniques varied for small canopy gaps (25 to 50cm), but were very similar for moderate (50 to 200cm) and large (>200cm) canopy gaps. Our results demonstrate that canopy-gap size distributions can be reliably estimated from HR imagery in a variety of plant community types. Additionally, we suggest that classification goodness-of-fit measures are a potentially useful tool for identifying and screening out plots where precision of estimates from imagery may be low. Our results are consistent with other research that has looked at the ability to derive monitoring indicators from HR imagery.