Errors resulting from magnification variations of a optical system are largely generated in three-dimensional shape measurements based on depth-from-focus. In the case of measuring the surface morphology of tiny objects based on DFF, images are acquired with a very small interval so that magnification changes can be minimized. However, the magnification variations are actually existed in the acquired images and so focus measures are wrongly or ambiguously extracted. In this paper, a methodology with linear magnification calibrations, based on DFF, is proposed to make more accurate measurement in surface morphology with high depth discontinuity, compared with previous ones. Several experiments show that the proposed method outperforms existing ones without magnification calibrations.