H. 264/MPEG-4 AVC features flexible macroblock mode decision

which makes the number of data bits produced by encoding header information in H. 264 video stream vary irregularly and hard to be predicted. As H. 264 encoder reduces the data bits for transform coefficients effectively

the number of header bits takes a larger portion in the overall bits than that of any former video compression standard. Thus precise estimation of header bits can effectively improve the performance of the H. 264 rate control algorithm. In this paper

we apply the Kalman Filtering theory to improve the precision of estimating the number of the header bits at the frame layer

and incorporate the estimation model to frame layer rate control.