Most existing rate control schemes in the literature calculate quantization parameters of the macro-blocks (MB) based on the standard deviation of the residue before quantization. The probability of non-zero coefficients after quantization is a new factor that has an interesting property. The rate tends to be linear with this probability. In this work, we design the rate control based on this special factor with its linear property. This work investigates the relationships of bit rate and distortion with MB quantization step size and its interesting probability. We establish linear rate and linear distortion models and propose a rate control scheme based on these models. The proposed algorithm has a low quantization overhead, low computation complexity and is independent of the distortion factors, compared with TMN8 rate control. The experimental results suggest that our scheme can achieve PSNR gain over TMN8.