In Scalable Video Coding (SVC) the coding efficiency for hierarchical prediction structures is highly dependent on how to choose the quantization parameters (QP) for pictures of different temporal layers. For pictures that have strong temporal correlation and pictures at low bitrate, large peak SNR (PSNR) fluctuations will be produced when using the strategy presented in the proposal JVT-P014. In order to decrease the fluctuations this paper proposed a new method which combines the persistence of vision to decrease the fluctuations of the video quality. Simulated results show that the proposed technique is efficient. When compared with the method in the proposal JVT-P014, the PSNR difference decreases is about 20%, and it is more effectively for the pictures with strong temporal correlation. The algorithm in this paper provides valuable references for further study of optimizing encoder.