In video encoder, calculating speed of hard decision quantization (HDQ) is fast, however, the rate distortion performance is lost. Comparatively, the coding performance of best soft decision quantization (SDQ) is better but the calculation is complex. In order to solve the limitation of these two algorithms, this paper proposes a quantization algorithm of memory source model based on HDQ. The contributions of this paper as follows. Firstly, we puts forward to a framework of hard decision quantization algorithm based on adaptive threshold which takes rate consumption into account, heuristic threshold modeling. Secondly, give the method which can determine the model of rate threshold offline based on the principle of maximizing the probability of right judgment and minimizing the probability of wrong judgment. Finally, according to the coefficient of quantization position and the distribution of quantization parameter, an adaptive threshold model is constructed based on rate distortion optimization, and the optimal parameters of the model are determined. Experiments show that compared with fixed offset HDQ algorithm, HDQ with adaptive threshold based on source memory in this paper obtains significant performance improvements, BD-PSNR has 0.0964dB boost, BD-Rate has about 3.5723% bit rate savings. Compared to SDQ, the additional computational complexity of this paper is low in real time computation. Adaptive HDQ based on this model is very suitable for designing of hardware encoder architecture.