A subsampling-predictor associated approach for fast global motion estimation

Global motion estimation (gme) has many important roles in numerous applications, such asvideo compression, image stabilization, video-object segmentation, and etc. one well-knowngmemethodis the gradient-based technique. this method uses optimization techniques, like the levenbergmarquardtalgorithm, to minimize estimation error. such algorithms require an initial value for the initializing step.in this paper, we propose a simple and reliable gme structure with a new predictor. this structure uses athree-step search and a predictor for the initializing step. it is also incorporated with a fast gme methodthat uses pixel subsampling. this incorporation reduces the computational complexity of gme without asignificant loss of accuracy. this structure has less computational complexity and similar accuracy versuscommon methods.