Standard MRI cine exams for the study of cardiac function are segmented over several heartbeats and thus require a breath-hold to minimize breathing motion artifacts, which is a current limitation of this approach. The purpose of this study was to develop a method for the measurement and correction of respiratory motion that is compatible with cine imaging. Real-time images were used to measure the respiratory motion of heart, to allow translations, rotations, and shears to be measured and corrected in the k-space domain prior to a final gated-segmented reconstruction, using the same data for both purposes. A method for data rejection to address the effects of through-plane motion and complex deformations is described (respiratory gating). A radial k-space trajectory was used in this study to allow direct reconstruction of undersampled real-time images, although the techniques presented are applicable with Cartesian k-space trajectories. Corrected and uncorrected free-breathing gated-segmented images acquired over 18 sec were compared to the current standard breath-hold Cartesian images using both quantitative sharpness profiles (mm(-1)) and clinical scoring (1 to 5 scale, 3: clinically acceptable). Free-breathing, free-breathing corrected, and breath-hold images had average sharpness values of 0.23 +/- 0.04, 0.38 +/- 0.04, and 0.44 +/- 0.04 mm(-1) measured at the blood-endocardium interface, and clinical scores of 2.2 +/- 0.5, 4.2 +/- 0.4, and 4.7 +/- 0.5, respectively.