Video stabilization based on optical flow and regression analysis

Abstract

Video stabilization is an important video enhancement technology which aims at removing annoying unwanted motion from videos which is being used by various vision applications as a pre-processing activity. Since jitter levels cannot always be fully determined by systems automatically without affecting to the desired camera movements, it is essential to provide an easy way for ordinary users to decidc what to be considered as jitter./ A major problem of current general purpose software video stabilizers is removing some of the desired motions or leave undesirable motions in video streams because of the incapability in determining what to be considered as wanted/unwanted. This dissertation describes a practical software based solution for video stabilization that address this by providing a better control over to the user in determining level of stabilization willing to have in the output video./ The proposed system initially determines to level of jitter automatically and provides high flexibility to the user to change the level of the jitter to be eliminated using few user parameters before generating full frame based stabilized output. This process is backed up by set of graph outputs that can be easily understood by ordinary user. Video is perceived as stable when the scene shown in consecutive frames has smooth flow over time. Based on this observation, proposed method removes horizontal and vertical undesirable movements while providing above flexibility using non-linear regression analysis based algorithm in iterative manner./ The proposed system initially determines to level of jitter automatically and provides high flexibility to the user to change the level of the jitter to be eliminated using few user parameters before generating full frame based stabilized output. This process is backed up by set of graph outputs that can be easily understood by ordinary user. Video is perceived as stable when the scene shown in consecutive frames has smooth flow over time. Based on this observation, proposed method removes horizontal and vertical undesirable movements while providing above flexibility using non-linear regression analysis based algorithm in iterative manner.