FPGA Implementation of Real Time Image Stabilization
In the latest technology, there are different applications which are working on videos. Particularly, videos from cameras which are placed on moving platform. But due to uneven surface or camera vibration, captured video tends to have undesired jitters, shakes and blurs. This results into unpleasant viewing experience and also affect on video processing application such as surveillance or some military application also. Digital video stabilization is an essential video improvement technology which is focusing on removing unnecessary camera vibrations from video sequences. In the proposed method, first feature points are detected with the Features from Accelerated Segment Test (FAST) corner detection algorithm and then these feature points are extracted using efficient Fast Retina Key point (FREAK) descriptor. Then based on hamming distance feature points are matched between consecutive frames. Next, the matched point pairs are fitted to the affine transformation model using a M-estimator SAmple Consensus (MSAC) algorithm, which is a variant of the RANSAC algorithm based approach to estimate inter-frame motion strongly. Then the calculated results are used to find out the cumulative motion parameters between the conjugative frames, and the translational components are smoothed by a Kalman filter representing intentional camera movement. Experimental results have shown that the proposed system can deal with standard recorded video input including arbitrate translation and rotation and can produce full-frame stabilized output providing a better viewing experience. We are trying to achieve real time video stabilization system by using pipelining and parallel processing strategies, designing the whole process using a novel complete fully pipelined architecture and implemented on FPGA.
Index terms - Feature detection, Feature extraction, motion estimation, video stabilization.