Multi frame Super Resolution Reconstruction (SRR) is an important problem in image processing. By using the Bayesian framework, the SRR model can be divided into two parts: the degradation model and the prior model. Nowadays, a great many of researchers focus on the prior models, and many excellent prior models have been proposed. However, the commonly used degradation model is an ideal and simple model. It just considers the noise error in the degradation process. However, there is much information lost in the degradation process. Thus in this paper, a novel degradation model is proposed, which aims to reconstruct a better high resolution image and the lost information. The Experimental results show that the proposed degradation model outperforms the existing degradation model.