Catheter/endoscope-based optical coherence tomography (OCT) is a powerful modality that visualizes structural information in luminal organs. Increases in OCT speed have reduced motion artifacts by enabling acquisition faster than or comparable to the time scales of physiological motion. However motion distortion remains a challenge because catheter/endoscope OCT imaging involves both circumferential and longitudinal scanning of tissue. This paper presents a novel image processing method to estimate and correct motion distortion in both the circumferential and longitudinal directions using a single en face image from a volumetric data set. The circumferential motion distortion is estimated and corrected using the en face image. Then longitudinal motion distortion is estimated and corrected using diversity of image features along the catheter pullback direction. Finally, the OCT volume is resampled and motion corrected. Results are presented on synthetic images and clinical OCT images of the human esophagus.