The Method for Adaptive Material Classification and Pseudo-Coloring of the Baggage X-Ray Images

Krzysztof Dmitruk, Marcin Denkowski, Pawel Mikolajczak, Emil Benedykciuk

Abstract: Baggage X-ray scanners are one of the most widely used tools for maintaining mass security. Baggage scanners use the same operating principle as their medical counterparts, but the task entrusted to the scanner operator is different from that of the doctor. The scanner operator’s task is to find if there are any dangerous objects in the X-ray image. The operator has to evaluate the shape and kind of material of the scanned objects within a few seconds. Therefore, there is a need for algorithms and methods that analyse such images. This paper presents the dual-energy X-ray scan image segmentation algorithm Adaptive Horizontal Material Classification (AHMC) that classifies materials into multiple classes. The effect is obtained by clusterization of the two-dimensional histograms of low and high energy images using the sliding window method. On the basis of those histograms, local material classes are created. As a result, local classes are combined into global ones, corresponding to the specific material. Our experiments show that the proposed method achieves performance on the same level in comparison to the standard semi-automatic lookup table based methods, but due to its ability to create any number of material classes that every object in an image is made of, outperforms these methods and act as an initial instance segmentation. The generated segmentation is then used in an innovative pseudo-colorization algorithm for X-ray scans.