Predicting Detection Performance on Security X-Ray Images as a Function Of Image Quality
Developing methods to predict how image quality affects task performance is a topic of great interest in many applications. While such studies have been performed in the medical imaging community, little work has been reported in the security X-ray imaging literature. In this work, we develop models that predict the effect of image quality on the detection of improvised explosive device (IED) components by bomb technicians in images taken using portable X-ray systems. Using a newly developed NIST-LIVE X-Ray Task Performance Database, we created a set of objective algorithms that predict bomb technician detection performance based on measures of image quality. Our basic measures are traditional Image Quality (NSS)-based measures that have been extensively used in visible light (VL) image quality prediction algorithms. We show that these measures are able to quantify the perceptual severity of degradations and can predict the performance of expert bomb technicians to identify threats. Combining NSS- and IQI-based measures yields even better task performance prediction than either of these methods independently. We also developed a new suite of statistical task prediction models that we refer to as Quality Inspectors of X-ray images (QUIX), which we believe to be the first NSS-based model for security X-ray images. We also show that QUIX can be used to reliably predict conventional IQI metric values on distorted X-ray images.
Portable transmission X-ray imaging systems are used by military and civilian bomb technicians to screen suspicious packages and objects for explosives, bombs and other threat items contraband. Their easy deployment and high detection efﬁciency makes them ideal for screening of hard-to accessible places. The quality of the X-ray images captured by these systems serves as an important indicator of the manufacturing quality and overall performance of the imaging system. Several intrinsic and extrinsic factors affect the quality of X-ray images. The geometry of a portable X-ray imaging device, such as the size of detector photo sensors and the generator’s focal spot, has a strong inﬂuence on the quality of capturedX-rayimage.Photon-limitednoiseduetotheinherent variation of photon inﬂux at each photo sensor is a major source of noise in X-ray images
There exist internationally standardized methods for objectively measuring the quality of images produced by portable transmission X-ray systems. These objective quality metrics, which we will refer to as image quality indicators (IQIs), operate by making specific quantitative measurements on images of standard test objects obtained under highly specific test conditions. IEEE/ANSI N42.551 includes a detailed description of the measurement and performance requirements of these conventional IQIs, which include ‘Useful penetration’, ‘Organic material detection’, ‘Spatial resolution’, ‘Dynamic range’, ‘Noise’, ‘Flatness of field’, ‘Image extent’, ‘Image area’, and ‘Aspect ratio’. While these IQIs do provide reliable measurements of image quality, their computation also involves the use of precisely defined test objects that are imaged under strictly defined laboratory conditions, which consumes significant amounts of time, cost and effort.
The Gaussian Scale Mixture (GSM) model provides a robust description of the statistics of band pass wavelet coefficients of natural VL images and, as it turns out, of X-ray images as well. It has been successfully applied to numerous perception driven image processing applications. Recently, a generalized Gaussian scale mixture (GGSM) model was proposed to model the band pass statistics of distorted VL images , and shown to better represent the statistics of both pristine as well as distorted VL images than the GSM model. To demonstrate this, assume that an X-ray image (distorted or not) has been subjected to a band pass process such as a wavelet filter. The GGSM model of the marginal distributions of band pass VL (and X-ray) image coefficients are heavy tailed, reflecting the property that natural images are predominantly smooth with sparsely distributed singular structures.
Predicting Detection Performance on Security X-Ray Images
- all of which were either current or former bomb technicians. Each subject viewed an average of 20 X-ray images ranging from a minimum of 5 to a maximum of 39 images.
- Considering the high proficiency and expertise of the subjects, we presented each image to an average of only 2.27 subjects.
- The size of the database was limited by geography and the availability of this small and specialized population.
- They were asked to locate and identify any potential IED components and annotate the image by drawing a box around it using a mouse.
- human subject responses on a sample of X-ray images, along with baseline annotations of all relevant IED components.
- One type of band pass X-ray image decomposition we use is a steerable pyramid along 2 scales and 2 orientations (vertical and horizontal).
- To characterize wavelet coefficients as a GGSM vector, we utilize the neighborhood structure in of 27 coefficients: 25 from the same sub band (the nearest 5 _ 5 neighbors),
- In this work, we contribute to solve this problem by analyzing the detection performance of expert observers on distorted X-ray images, and we build perceptual X-ray image quality models that reliably predict observers’ task performance.
- While a considerable vein of research in this direction has been developed in the medical imaging field, only a little work has focused on the visual task performance in security X-ray imaging
- . In several factors, including sensitivity and response time, were studied on visual scanning and target detection tasks, where human observers were tasked with searching for a knife inserted at randomly different angles in chromatic X-ray images of cluttered baggages.
- In another work, participants were asked to identify improvised explosive devices (IEDs) in a brief presentation of suspicious baggages, and a model observer developed for a different medical imaging task was adapted to explain the observers’ performance.
- To the best of our knowledge, there is no reported work that deals with task-based image quality assessment (IQA) of security relevant X-ray images.
- MATLAB 7.5 and above versions
0.00 average based on 0 ratings
More Things You Might Like This
Abstract: Although the educational level of the Portuguese population has improved in the last decades, the statistics keep Portugal at Europe’s tail end due to its high student failure rates. In particular, lack of success in the core classes of Mathematics and the Portuguese language is extremely serious. On the other hand, the fields of
Abstract: Advances in natural language processing (NLP) and educational technology, as well as the availability of unprecedented amounts of educationally-relevant text and speech data, have led to an increasing interest in using NLP to address the needs of teachers and students. Educational applications differ in many ways, however, from the types of applications for which
Machine Learning based Regression Model for Prediction of Soil Surface Humidity over Moderately Vegetated Fields
Abstract: Agriculture is one of the major revenue producing sectors of India and a source of survival. Numerous seasonal, economic and biological patterns influence the crop production but unpredictable changes in these patterns lead to a great loss to farmers. These risks can be reduced when suitable approaches are employed on data related to soil