Recovering the three dimensional structure of a scene accurately and robustly is important for object modelling and robotic grasp planning, which in turn are essential prerequisites for grasping unknown objects in a cluttered environment. Shape recovery techniques are broadly described as either passive or active. Passive methods include recovering shape from a single image using cues such as shading, texture or focus, and shape from multiple views using stereopsis or structure-from-motion. Passive shape recovery has relatively low power requirements, is non-destructive and more akin to our biological sensing modalities. However, the accuracy and reliability of passive techniques is critically dependent on the presence of sufficient image features and the absence of distractions such as reflections. Active sensing is necessary to achieve the accuracy required for the object modelling and tracking techniques described in the following chapters.