Two images of the same place: one taken in the morning from a drone, the other in the evening from a satellite. The first shows shadows from trees, the second shows glints on the water. It's like two accounts of the same event in different dialects: the words are similar, but the grammar is different. Image registration is a translator that finds common ground between such frames.
The translator doesn't just match words; it looks for anchor points. In images, these are building corners, river bends, field contours. First, it detects these features, then matches them across frames, estimates how one image needs to be shifted, rotated, or stretched to align with the other, and finally applies that transformation. In this way, two "dialects" become one text.
But there is no universal translator. For rigid shifts and rotations, external markers or surfaces work well; for quick estimates, moments and principal axes; for complex distortions, soft computing methods: neural networks, fuzzy sets, evolutionary algorithms. Each method is its own dialect, and the choice depends on how "different the languages" of the images are.
Previously, researchers had scattered methods without an overall map. Now a classification has emerged based on ten criteria, from dimensionality to modality. It's like a dictionary that specifies the best translator for each pair of dialects. In medicine, such a translator combines anatomy and function; in ecology, it merges forest images from different years; in computer vision, it aligns frames for motion tracking.
But even the best translator sometimes stumbles. When an image is covered by clouds or an object is partially hidden, anchor points are lost. And then two accounts of the same place remain untranslated. The task of automatic registration under such disturbances is an open question that engineers and scientists are grappling with.
For aligning MRI and PET brain images, mutual information is often used: it finds common patterns in intensities, even if one image shows anatomy and the other shows metabolic activity.