In previous study, a tomographic imaging system is built up for measuring the three-dimensional refractive index distribution inside the micrometer-sized biological cell by optically driven full-angle rotation scheme based on digital holographic microscopy, named as optical-driven tomographic DHM (OT-DHM) system. However, a small perturbation of the system will lead the inaccurate of the positions and the orientation of the micrometer-sized sample, thus the automatic calibration of the reconstructed phase images in the OT-DHM system is required. For this purpose, a novel model-based algorithm is proposed, in which we employ a 3-D ellipse shape for modeling the samples. The parameters of the ellipse-like shape on a small number of the projections are estimated and used them to build up the 3-D ellipse model of the samples. In advance, the reconstructed phase images are highly contaminated by the uneven background and coherence speckle noise. The block-based between-class criterion is used to suppress the effect of the non-uniform background, and the anisotropic diffusion process is utilized for the noise cleaning, including shot noise and speckles noise on the reconstructed phase. The boundary of the cell in each projection can be considered as the 2-D ellipse, and used to estimate the parameters of the 2-D ellipse. The established 3-D ellipse shape is applied for the calibration of the spatial positions and the orientations of the all other rotational angles. With the automatic calibration algorithm, the OT-DHM system can effectively reconstructed the three-dimensional refractive index distribution inside the micrometer-sized samples.