Image registration is an important pre-processing step prior to detecting changes using image data and is increasingly accomplished using automated methods. Differences in illumination between images being registered can pose a challenge to obtaining accurate image registration results using automated routines. In high spatial resolution imagery, shadows represent a major source of the illumination variation that can occur between time sequential images. The connection between shadows and registration accuracy is acknowledged in the literature but has not been studied in detail. Through this research I seek to determine if there is a statistically significant relationship between amount of shadow and image registration accuracy, and explore whether masking and normalizing shadows leads to improved automatic registration results. High spatial resolution aerial image pairs captured from approximately at the same location in the sky through an approach called Repeat Station Imaging, were automatically co-registered using a software called SIFT and RANSAC Alignment (SARA) and their registration accuracy was compared the amount of total shadow and shadow movement present between their images. A direct and significant relationship (P < 0.01) was found between the magnitude of co-registration error and both the amount of total and transient (shifting) shadow in the pairs. Shadow normalization was explored using linear correction, gamma correction, and histogram matching approaches. Linear correction facilitated the most accurate co-registration results of the three and was compared against the original image pairs as well as versions whose shadows are masked. Masking shadows did not improve automatic co-registration accuracy in any of the image pairs. Normalizing shadows improved automatic co-registration results in image pairs where there is a substantial amount of shadow movement, but not in pairs that experience little to no shadow movement.