Image segmentation of concealed objects detected by passive millimeter wave imaging

We address an image segmentation method to detect concealed objects by a passive millimeter wave (PMMW) sensor. The PMMW imaging can create interpretable imagery in low-visibility conditions. Indeed, the millimeter waves penetrate into textile materials such as clothing and hair, and the reflectivity on the metal and man-made objects is high, therefore concealed objects can be easily detected by a PMMW imaging system. We develop automatic thresholding methods for the segmentation of concealed objects detected by PMMW imaging. The automatic thresholding method is useful for the image with bimodal distributions. We compare several thresholding methods for three MMW channel images which have 8 mm-horizontal, 8 mm-vertical, and 3 mm-horizontal polarization, respectively. The experimental results show that concealed weapon region is well segmented from the background body by the presented methods.