The applicability of the simplified Gaussian gray diffusion model needs to be further verified theoretically, whose error analysis method relies on the analytical comparison with the typical traditional one. The two models are normalized respectively, a test pixel is chosen, and the evaluating criterion for the simplified model is established by analyzing the relative error of the normalized gray value regarding the test pixel. Star image simulation is conducted, and four relative error curves to the mapping position deviation from the pixel central coordinate are plotted, each of which represents a different case of Gaussian radius σ, manifesting that error decreases as σ increases as a whole. Gray-weighted centroiding method is carried out to the two series of simulated star images generated respectively by the two models. Centroiding absolute error decreases as σ increases, absolute error resulted from simplified model is about 2 orders of magnitude higher than that of the traditional one. Under the conditions of no noise and σ＝0.671, the maximum absolute error from the simplified model is only 0.033 pixel. Simulated results show that the simplified model is no longer applicable when Gaussian radius σ is rather small in terms of shape simulation of image point. Considering that latter algorithms of a star sensor, such as star matching and attitude determination, whose focus is just centroid, the simplified model still possesses applicability due to its advantage of small calculating amount, also because its error magnitude is small enough to be neglected anyway.