Total variation regularization algorithm for video stabilization in a digital camera

This paper presents a motion modeling technique for digital video stabilization system. Our modeling approach may be applied for separation of the intentional camera motion and undesired shaky motion from estimated motion vector. The technique used for this separation is called total variation regularization technique and we believe that this motion modeling technique is well-suited for removal of unintentional motion noise. The experimental results establish the fact that our proposed methodology improves the quality of video stabilization.