Camera Motion Modeling for Video Stabilization Performance Assessment
M. J. Tanakian, Mehdi Rezaei, Farahnaz Mohanna · 2011
Digital video stabilization (DVS) allows acquiring video sequences without disturbing jerkiness, removing unwanted camera movements. A good DVS algorithm should only remove the unwanted motion while maintain the intentional motion. Numerical performance assessment of motion compensation (MC) unit of a digital stabilizer is a difficult task since the pure intentional camera motion path as reference is not available. To solve the lack of reference, we propose a model for the intentional camera motion to produce synthetic camera motion path as reference. Having a reference path for intentional camera motion, a random noise is added to it to simulate the unwanted camera motion. When provided motion path is processed by a DVS algorithm, computing a distance measure such as mean-square error (MSE) between the processed motion path and the reference path can be considered as a numerical performance measure.