As the rate control model in H.264/AVC ignores the influence of video features on the selection of a Initial Quantization Parameter (QP0), this paper proposes a novel QP0 algorithm based on video features to improve the performance of H.264/AVC rate control. First, video features which have influence on the determination of QP0 were analyzed, such as (bit per pix, bpp), video complexities and length of Group of Picture (GOP). Then, functional relationships between QP0 and bpp, QP0 and video complexity were established through extensive simulation. Finally, considering the influence of the length of GOP on the QP0, the QP0 model was revised. Experimental results show that the proposed algorithm increases the average Peak Signal-to-Noise Ratio(PSNR) of reconstructed video by 0.185 dB for Quarter Common Interchange format(QCIF) and 0.144 dB for Common Interchange format (CIF) sequence as compared with the algorithm in JM12.2. According to the error of rate control, the control amplitude is increased to 37.3% for the QCIF sequence and 11.2% for the CIF sequence. Moreover, the PSNR fluctuations are reduced by 50% for both sequences. In conclusion, the proposed method not only decreases the error of rate control, but also depresses PSNR fluctuations effectively.