Summary: | 碩士 === 義守大學 === 資訊工程學系 === 90 === Motion compensation method compresses a digital video sequence by removing temporal redundancy between frames. It predicts the current frame based on the previous frame and the motion information between the two frames. Moving objects cause local intensities changes whereas the changes due to the motion of camera are global. Since the global motion would increase the number of moving blocks, many approaches for eliminating the global motion have been developed recently.
This thesis presents an efficient method to estimate global motion parameters. We use an affine model to represent the global motion and develop a recursive least square scheme to speed up the estimate of motion parameters. In addition, in the proposed method the motion vectors in the background area are selected for the estimate of affine parameters, which improves the estimate reliability.
The simulation results indicate that the proposal method improves both computational efficiency and estimation accuracy, as compared to the iterative least square method.
|