摘要 |
A video stabilization method includes detecting a plurality of feature points of an inputted image frame. Next, a block motion estimation is performed onto the inputted image frame according to the feature points, thereby generating a plurality of block motion vectors corresponding to the feature points. Then, the block motion vectors are converted to a histogram to perform a statistical process. Thereafter, an initial static background motion vector is determined according to a statistic value of the histogram. Thereafter, a Gaussian distribution model and a Maximum a Posteriori (MAP) estimation are adopted to correct the initial static background motion vector to generate a global motion vector. Then, a smoothing process is performed onto the global motion vector to generate a smoothed global motion vector. A video stabilization system is also provided here.
|