I. Introduction
Structure from small motion (SfSM) has recently attracted much attention due to its applicability to various tasks such as depth-based image editing, digital refocusing, etc. One of the main goals of SfSM is to reconstruct the three-dimensional (3-D) structure of a scene acquired with a portable, everyday camera such as a smartphone. When taking a picture, the camera usually undergoes slight movement due to the user's natural motion, such as handshaking, breathing, and heart-beating. In SfSM, such natural small motion is taken as advantageous information for 3-D reconstruction simply by taking a short video, so called small motion clip [1], at that moment.