In recent years, online purchasing of clothes and shoes has become increasingly common. Although this is convenient, it can be difficult to choose the correct shoe size. While 3D foot scanners can accurately measure foot size and shape, this expensive and large scale equipment is not generally accessible for personal use, and there is a need for some simple and accurate means of measuring the foot in 3D. Recently developed smartphones with depth cameras enable easier measurement of 3D shapes, and this paper describes a method for measuring foot shape using a 3D point cloud captured from multiple directions by such a camera. As a 3D point cloud can potentially include noise or may omit occluded parts of the foot, we propose the use of a dataset of 3D foot shapes collected by a precise 3D shape scanner. We show how a deformable model can be generated by performing a principal component analysis on this dataset, minimizing error to recover a complete and high-accuracy 3D profile of the entire foot. We tested this method by comparing the 3D shape so produced to the 3D shape measured by the 3D scanner. The proposed method was found to scan foot shape with an error of about 1.13 mm. As demonstrated experimentally, the contribution of our work is in introducing the deformable model of 3D foot shapes based on principal component analysis, so that accurate shape models can be calculated from noisy and occluded 3D point clouds obtained via smartphone input.