Detection and segmentation of concealed objects in terahertz images
Xilin Shen, Charles Dietlein, Erich N. Grossman, Zoya Popovic, Francois Meyer
Terahertz imaging has been shown to successfully detect objects concealed underneath clothing, by measuring the radiometric temperatures of different objects on a human subject. The goal of this work is to automatically detect and segment concealed objects in broadband 0.1-1THz images. Due to the inherent physical properties of passive terahertz imaging and associated hardware, the images have low contrast and low signal to noise ratio (SNR), and therefore noise reduction for the raw image data will help improve the performance of the object detection. Passive terahertz images are represented through radiometric temperature, and we assume that the data can be modelled by a piecewise smooth function, allowing effective application of the anisotropic diffusion algorithm. The low contrast of terahertz images poses a problem for existing segmentation methods, and this paper presents a solution using an approach referred to as Multilevel Thresholding. This method combines the analysis of the image histogram and the geometry of the intensity isocontours. Two state-of-the-art unsupervised methods are applied to the images and fail to identify the concealed object, while the method presented in this paper correctly detects the object. In addition, the results of our approach compare favorably with an existing state-ofthe- art supervised image segmentation algorithm.
, Dietlein, C.
, Grossman, E.
, Popovic, Z.
and Meyer, F.
Detection and segmentation of concealed objects in terahertz images, IEEE Transactions on Image Processing, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=32839
(Accessed December 3, 2023)