Corresponding Points

Two images, the same spatial position — except you have to find them anew in each frame. Corresponding points are the tools with which tracking software and motion matching algorithms understand how a camera is moving or how a 3D object is changing its position. You mark a characteristic point in frame 1 (a corner, a scratch, an irregularity on a surface) and the same point in frame 2 — the software then calculates the movement from the displacement between these positions.

On set, you need at least four corresponding points per shot to enable stable 3D reconstruction. Why four? Three points only define a plane; the fourth gives you depth. The more points you have — five, six, or more — the more robust the tracking becomes. The best strategy is to choose points distributed across the entire image space: top left, top right, bottom left, bottom right. Avoid arranging them in a line, otherwise the software loses spatial reference.

In practice, you place these points either manually — with marker pens on objects, with taped-on dots in the scene, or directly in the tracking software interface — or you let the software detect them automatically. The automatic process works well with high contrast and stable features (sharp edges, textures). It becomes critical with flat walls or moving water — then you have to help by hand. Some DoPs tape reflective tracking markers (so-called tracking balls) into the image to make the algorithm's job easier.

A common mistake: choosing corresponding points that change during the shot — a water surface, a fluttering cloth, an actor. Stable, stationary, or predictably moving points are the gold standard. For motion matching in camera tracking, the software works iteratively — it starts with your manual points and refines the position automatically over multiple frames. That's why the initial placement is so crucial: incorrect starting corresponding points lead to drift and unstable tracks.