Humans are extraordinarily adept at identifying features in, and extracting information from, images. Despite the untold billions of dollars invested in automated techniques, the most import image recognition tasks, whether it be scanning luggage before it is put on a plane, reading out medical radiographs, or identifying targets for a military strike – all are still carried out by humans. There is a rich body of research, still ongoing, that seeks to understand how humans understand information in imagery. In this section, we don’t attempt to cover all of the theories behind this research, but we do consider two alternative explanations.
One theory is referred to as scene gist. The underlying principal of scene gist is that humans develop an understanding of an image within a fraction of a second without having to identify all of the objects in that scene. An alternative theory is that there is a set of fundamental building blocks, called geons, which humans virtually combine to recognize objects. Scene gist can be thought of a top-down approach, whereas geons are a bottom-up approach. The reality is that an experienced image interpreter uses elements of both of these. On one hand, a forest may be easily identified in a split second. The interpreter will then use this information to identify other characteristics such as species or condition. On the other hand, distinguishing between a real and fake transporter erector launcher (TEL) may require the image analyst to break down the individual components to look for discrepancies.