Neural Network based Shape Recovery from SEM Images Using Secondary Electron Image and Reflecting Electron Image

Scanning Electron Microscope (SEM) is used to see some object with micro size and shape. Although SEM has been a popular measuring equipment, some limitation exists to recover the 3D shape of an object with complicated shape especially for an object with some convex and concave shape. SEM uses the secondary electron microscope in general but some SEM can use the reflecting electron to see the object image further. When the second electron image is used to recover the shape, only one image is available to recover the shape. However, there are some limitations to recover the shape in general. So this paper proposes a neural network based approach to recover 3D shape of an object with some complicated shape including the object without any texture information using not only the secondary electron image but also the reflecting electron image. Experiments are demonstrated with these SEM images to evaluate the validity of proposed approach.