FAST Based Video Stabilization with Preserved Intentional Motion and Smear Removal

Visual sequence recording device may have unintentional jitter motion due to the unwanted movements in camera platform or hand motions. This unintentional motion causes video instability. These unstabilized videos have poor quality and jittery frame motion, hence not suitable for viewing or any other post processing applications like visual tracking etc. Thus, digital video stabilization is of prime importance. A fast algorithm is proposed in this paper for digital video stabilization. Algorithm is also capable of preserving intentional camera motion intelligently and removes smear introduced by unintentional camera motions. Global Motion Vector (GMV) is estimated using Features from Accelerated Segment Test (FAST). Proposed video processing algorithm considers only two number of frames at a time for computation and compensation of global camera motion, which results in fast execution. Hundred strongest robust features are tracked for video stabilization against any translation as well as rotational disturbances, resulting in robust online stabilization. Performance of the proposed algorithm is evaluated on a number of videos w.r.t. ground truth and output stabilized video has improved quality factor of up to 18 dB.