This paper proposes a global motion compensated saliency estimation method with a hand-held camera for video retiming. To estimate the saliency score of only the foreground objects in the case of a moving camera, the foreground object mask is estimated by detecting contiguous outliers in the low-rank representation. Motion features of the object are computed based on the optical flow, and then global motion compensated motion features are obtained using the foreground mask. Finally, the global motion compensated saliency scores are estimated using motion-based features, and input video is globally retimed, where the output video is played slowly at frames with high saliency scores. Experimental results show that the proposed method estimates more accurately the saliency scores in the case of a moving camera.