Advanced Image Processing for Rapid Threat Object Identification in Terahertz Images

With an increasing trend of anti-social activities and terrorist attacks, there is a persistent demand for rapid and effective security screening systems, particularly in crowded public places such as in metros stations, airports, etc. Nowadays, terahertz (THz) based rapid scanners and cameras are employed as an alternative to the traditional X-ray baggage scanners due to the ability of THz waves to penetrate most nonmetallic materials and low ionizing radiation. Although such rapid THz line scanners and cameras are commercially available, the quality of the THz images acquired with these devices are very poor. This requires advanced image processing techniques to be employed for rapid threat object detection and identification. In this work, a sequence of image processing steps has been adopted which include (i) Weiner filtering, (ii) Lucy-Richardson deconvolution with the appropriate point-spread function, (iii) histogram based global thresholding of the image. This sequence of image processing operations has resulted in refined images thereby enabling rapid threat object detection and identification. This approach may be utilized as a preprocessing step before implementing sophisticated neural networks for automatic threat object detection and classification.