Researchers developed a new method to find important features in images. They used data from people and machine learning to figure out which features are most noticeable. The results showed that people focus on specific features when judging similarity between images, and tend to exaggerate the differences between these features.