Most of the internet shoe shops are interested in selling rate increasing. Main problem is inside too high percentage of delivery returns due to wrong shoe fit. Minimization of wrong shoe size selection can be achieved using virtual shoe fitting. In this scheme customer (private person without specific measuring equipment) should perform a measurement of some important (non-trivial) dimensions of his(her) feet and pass them through on-line service to the online store. The disadvantage of this approach is the lack of accuracy of measurements with the measuring instruments available to the client. Using digital foot images it is possible to detect characteristic foot sizes automatically, with enough high level of accuracy. This paper proposes a method for horizontal and vertical foot contours restoration from digital images, and detection physical foot size with sufficient accuracy for shoe fitting applications.