A composite image is produced for use in HYPR processing a current image frame. The amount of a priori data used to form the composite is determined by the amount of subject motion. The current composite image may be spatially registered with the current image frame to offset subject motion before being used to form an updated composite image. Subject motion may be analyzed on a frame-by-frame basis, a region-by-region basis or a pixel-by-pixel basis to optimize the SNR of the updated composite image.