The motion estimated from tennis video is unable to analyze player's true movement due to the diagonally located camera and the high noises.In this paper

a transform algorithm of Motion Vector Field (MVF) is proposed to revise the deformed motion vector based on the pinhole camera model.In this algorithm

the techniques of foreground object mask and global motion compensation are incorporated to reduce noises.Previous methods based on MVF analysis are further utilized to recognize basic actions of either tennis player.And

the experiments on live broadcasted tennis videos show very encouraging results.