Detection and Segmentation of Concealed Objects in X-Ray Compton Backscatter Images

A fast and effective double-level segmentation method is proposed to detect and segment the concealed objects adaptively in various real-time X-ray Compton backscatter (CBS) images. The double-level segmentation method is composed of the global segmentation and the local segmentation. The global segmentation separates the body area from background area on the basis of a two-class Gaussian density, and the local segmentation segments the concealed objects from the body area based on k-means algorithm and a two-class mixture of Gaussian densities. We have compared our method with one multilevel segmentation method. Experiments confirm our method more efficient and effective in detection and segmentation.