When performing relative radiometric normalization (RRN) on higher spatial resolution images using a reference image with lower spatial resolution, it is common to downsample the high-resolution (HR) target image to match the spatial resolution of the low-resolution (LR) reference image for the transfer function estimation. In this study, we downsampled HR target images by pixel aggregation using different scale factors, generated transfer functions with LR reference images, and compared the similarity between the corrected images and the reference images. We examined two commonly used RRN methods that do not require pixel pairing: moment matching (MM) and histogram matching (HM). The evaluation metrics used for the comparison included the root mean square error (RMSE), the cumulative distribution function distance (CDFD) and the structural similarity index (SSIM). We found that variations in the scale factor can impact the accuracy of the multi-resolution image RRN. There exists an optimal scale factor value that results in the corrected high-resolution image being closest to the low-resolution reference image. Moreover, this optimal value changes with different evaluation metrics. Using the optimal scale factor does not increase computational costs, but does lead to improved correction results.