Modeling human visual detection of low-contrast objects in fluoroscopy image sequences