Automatic X-ray image segmentation for threat detection

Multithresholding and data clustering techniques are used to segment X-ray images for low intensity threat detection in carry-on luggage. The widely used statistical validity indexes methods do not generate a reasonable estimation of the optimal number of clusters and produce a biased evaluation of the segmented images acquired by different segmentation methods. We propose a method based on the Radon transform to determine the optimal number of clusters and to evaluate the segmented images. The method utilizes both statistical and spatial information from the image and is computationally efficient. Experimental results show that the proposed method produces results consistent with human visual assessment.