In this paper, based on MPEG-2 Test Model 5 (TM5) bit rate control algorithm, there are some improvements proposed: ① an intelligent integral action is proposed for improving the control precision of bit rate; ② after using DCT AC coefficients to define the measure of macroblock and picture coding difficulty, an improvement on target bits allocation is proposed for the decoded picture quality and conformance including scene change, which can adjust the target bits allocation adaptively by their measure. Simulation results show that it can achieve high control precision by using intelligent integral action, its relative error is reduced one order of magnitude compared with TM5; using the improvement on target bits allocation, the average decoding PSNR is increased visibly, and the decoded picture quality is improved when scene changes.