We herein describe several practical improvements to the method of obtaining the size distribution of magnetic nanoparticles (MNPs) using the magnetization curves obtained from a water-based MNP sample. A vibration sample magnetometer (VSM) was used in place of a superconducting quantum interference device (SQUID), to reduce MNP agglomeration and sedimentation by means of vibration during the measurements. A signal preprocessing method using background rejection was applied to improve the measurement technique for lower concentrations of water-based MNPs, because of the significance of the diamagnetic effect of water in dilute solutions. We further present the results of detailed studies using singular value decomposition (SVD) and the Tikhonov-regulated SVD method to reconstruct the size distribution of the original sample. The experiments show that the controlled use of SVD produces consistent results. Artificial oscillation found using both SVD and the Tikhonov-SVD methods indicates that the characterization of MNP using a magnetization curve could be subject to an upper limit to its accuracy for measuring size distribution.