An algorithm using fixed-lag smooth filtering to improve the precision of EIS algorithm is presented in this paper. Ftrstly, gray projection algorithm (GPA) is used to calculate the interframe global motion vectors. S...An algorithm using fixed-lag smooth filtering to improve the precision of EIS algorithm is presented in this paper. Ftrstly, gray projection algorithm (GPA) is used to calculate the interframe global motion vectors. Secondly, the fixed-lag smooth filter is used to smoothen the motion track of the original video sequence, according to different motion models. At last, the original sequence is compensated by using the compensation vectors calculated from filtering. This algorithm has been proved superior in filtering precision by experiments of video sequences in different motion scenes. This algorithm also satisfies the real-time request.展开更多
In this paper, a new approach to H-infinity fixed-lag smoothing is developed by applying the innovation analysis theory. The smoother is derived by resorting to the augmentation state. However, being completely differ...In this paper, a new approach to H-infinity fixed-lag smoothing is developed by applying the innovation analysis theory. The smoother is derived by resorting to the augmentation state. However, being completely different from the previous work,the augmented state here is considered as just a theoretical mathematical tool for deriving the estimator. A distributed algorithm for the Riccati equation of the augmented system is presented. The calcuhtion of the estimator does not require any augmentation. The comparison of the computation costs between the new approach and previous work is made. The main technique applied in this paper is the re-organized innovation analysis in an indefinite space.展开更多
文摘An algorithm using fixed-lag smooth filtering to improve the precision of EIS algorithm is presented in this paper. Ftrstly, gray projection algorithm (GPA) is used to calculate the interframe global motion vectors. Secondly, the fixed-lag smooth filter is used to smoothen the motion track of the original video sequence, according to different motion models. At last, the original sequence is compensated by using the compensation vectors calculated from filtering. This algorithm has been proved superior in filtering precision by experiments of video sequences in different motion scenes. This algorithm also satisfies the real-time request.
基金The work ofthe first author was supported bythe National Nature Science Foundation of China (No .60174017,60574016) .
文摘In this paper, a new approach to H-infinity fixed-lag smoothing is developed by applying the innovation analysis theory. The smoother is derived by resorting to the augmentation state. However, being completely different from the previous work,the augmented state here is considered as just a theoretical mathematical tool for deriving the estimator. A distributed algorithm for the Riccati equation of the augmented system is presented. The calcuhtion of the estimator does not require any augmentation. The comparison of the computation costs between the new approach and previous work is made. The main technique applied in this paper is the re-organized innovation analysis in an indefinite space.
