PURPOSE: In respiratory self-navigated coronary MRA, the selection of a reference position may have a direct effect on image quality. While end-expiration is commonly used as reference, it may be ill defined in cases of irregular breathing. Here, an iterative self-navigation approach that operates without a reference position was implemented and tested in healthy volunteers and patients. METHODS: Data were acquired in 15 healthy volunteers and in 23 patients. Images obtained with end-expiratory self-navigation were compared with those obtained with the iterative approach that incorporates cross-correlation to iteratively minimize a global measure of respiratory displacement. Vessel sharpness, length, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were evaluated while differences in breathing patterns between the two sub-groups were assessed, too. RESULTS: Vessel sharpness and length were similar for both methods in healthy volunteers. In patients, a significant improvement in vessel sharpness and length was obtained using the iterative approach. SNR and CNR remained constant. While end-expiration was the most frequent respiratory phase in healthy volunteers (57.6 +/- 16.2%), intermediate respiratory phases (43.4 +/- 30.1%) were predominantly found in patients. CONCLUSION: An iterative approach to respiratory motion correction in self-navigation may lead to significant improvements in coronary MRA image quality in patients with a less consistent end-expiratory respiratory phase. Magn Reson Med, 2015. (c) 2015 Wiley Periodicals, Inc.