Predicting Detection Performance on Security X-Ray Images as a Function of Image Quality.

Area of Science:

• Security imaging
• Image quality assessment
• Human-computer interaction

Background:

• Predicting image quality's impact on task performance is crucial.
• Limited research exists in security X-ray imaging compared to medical imaging.

Purpose of the Study:

• Develop models to predict how X-ray image quality affects improvised explosive device component detection.
• Assess the performance of bomb technicians using portable X-ray systems.

Main Methods:

• Utilized the NIST-LIVE X-Ray Task Performance Database.
• Developed objective algorithms using traditional image quality indicators (IQIs) and natural scene statistics (NSS).
• Created a new suite of statistical models named Quality Inspectors of X-ray Images (QUIX).

Main Results:

• NSS- and IQI-based measures effectively quantify image degradations and predict bomb technician performance.
• Combining NSS and IQI measures improved task performance prediction accuracy.
• The QUIX model, the first NSS-based model for security X-ray images, reliably predicts conventional IQI values.

Conclusions:

• Objective image quality measures can predict human performance in security X-ray tasks.
• Combined NSS and IQI approaches offer superior prediction capabilities.
• The QUIX model provides a novel tool for assessing security X-ray image quality and performance.