Scatter Based Novel Imaging Systems

Abstract

Imaging is the process of reproduction of an object’s form. Image science is a multidisciplinary field concerned with the generation, collection, duplication, analysis, modification, and visualization of images.Traditionally, the image of an object is generated by gathering optical power reflected or absorbed by the object directly with a detector array in a one-to-one manner. For example, in optical photography, reflected or emitted photons from object are directly captured by optical elements (e.g. lens, mirror) and focused onto a detector array. However, in many scenarios where such direct imaging is not feasible, one is forced to gather information that is only indirectly related to the object. In some extreme cases, we are forced to work with only scattered photons from the object of interest. In this work, we explore two of these special cases and propose two novel indirect imaging (scattering-based) approaches operating at different wavelengths. The first indirect imaging approach we consider a novel non-line-of-sight (NLOS) imaging technique. Non-line-of-sight imaging refers to the imaging of an object that is not directly visible from the viewer/imaging perspective. A classic NLOS setup usually involves emission and collection of photons actively with time-of-flight measurements. To accurately measure a photon’s traveling time, most approaches employ expensive and exotic optical sources and detectors such as ultrafast laser and ultra high-resolution single photon avalanche detector (SPAD). In this work, we instead propose a novel passive NLOS imaging method that works with ambient light illumination and employs only commonly available imaging elements. We develop a light path order based image formation model, a learning-based model parameter estimation process, and successfully demonstrate instant passive NLOS imaging. We discussed the limitation of this technique and possible applications We then shift our focus to ultra-short wavelength or higher photon energy. In this part, we study a different type of imaging referred to as X-ray diffraction tomography (XRDT). X-ray diffraction tomography is a tomographic imaging of the coherent scattering profile (X-ray form factor) of an object and its material. Coherent scattering is one of three forms of photon interactions possible in the X-ray photon energy ranging from 10 kev to 200 kev. It occurs when X-ray photon energy is relatively small compared to the ionization energy of the atom. When a coherent scattering event happens, the photon does not have enough energy to liberate the electron from its bound state and no energy transfer occurs. Instead, the X-ray photon gains momentum from the micro-structure of the material and undergoes a change in its direction (scatter). Therefore, the scatter direction contains vital information of the micro-structure of the material and is unique to different materials. The scattering pattern is called momentum transfer function or X-ray form factor. It is considered as the golden standard in material classification. In this work, we explore the inherent sparsity in the X-ray form factors and develop a sparsity regularized reconstruction algorithm to perform maximum-a-posteriori estimation for X-ray diffraction tomography. We demonstrate the possibility of achieving higher energy resolution with limited number of photon, which allows our system to run on a lower signal-to-noise ratio.