Segmentation of handguns in dual energy X-ray imagery of passenger carry-on baggage

This paper considers three different methods for segmentation of the handguns from X-ray images of passenger carry-on baggage. (1) The simplest approach is a direct greylevel based segmentation in which a fixed absolute threshold and region growing are used to identify candidate regions of the image. This is a pixel-based approach which uses only the simplest region based information to validate a candidate area. It is however, computationally simple and highly reliable and robust. (2) The second method considered is the classical two label probabilistic relaxation labeling (PRL) technique. This method is a hybrid between a model-based segmentation and a pixel-based approach. (3) The third approach considered is a physical model in concert with significant line and edge determination applied on top of the PRL method. The method is computationally expensive. It may provide significant savings in later processing which will out weight the initial computational costs, however, this will require further augmentation of the object model