Digital Video Stabilization with Preserved Intentional Camera Motion and Smear Removal

Abstract

Videos captured through camera may have unwanted jitters and vibrations. Most of the video algorithms, available in literature stabilizes the video against all the disturbances. It may happen that user may be recording video using intentional hand motion or by moving the camera fitted on a vehicle with very slow motion to capture scene or target. Then this recorded image sequence will have both intentional and unintentional motion. Conventional video stabilization algorithm will reject this intentional motion also. In this paper, a computer vision method is presented to segregate unintentional and intentional motion (both translation and rotational). Then unintentional jitter motion is filtered out and compensated to get smooth image sequence. This high frequency jitter motion also introduces motion smear. The authors have also addressed this issue and developed the algorithm to remove/minimize the motion smear. A novel approach is presented to estimate motion vectors using Affine modeling and SURF features detection to stabilize the video. Point Spread Function is parameterized using estimated intentional global motion vector to remove motion smear simultaneously while stabilizing the video. Promising video results have been obtained for digital video stabilization keeping intentional motion and removing motion smear.