Projective Geometry: Points at Infinity Explained

Build a clean mental model of points at infinity, homogeneous coordinates, and projective transforms so you can reason about perspective, invariants like cross-ratio, and why parallel lines meeting is a feature, not a bug.

Parallel lines meeting sounds like a trick until you see what it buys you. In Euclidean geometry you keep adding exception rules. Lines intersect unless they are parallel. Constructions work unless a line misses. Proofs split into cases. The projective move is to stop treating parallelism as special. You add just enough new points so every pair of lines intersects, and the geometry becomes smoother, not stranger.

That single change also forces a new habit. You stop asking where a line hits another in meters and start asking what is incident with what. Once you lean into incidence, the weirdness of infinity becomes convenient.

What projective geometry keeps and forgets

Projective geometry is built to preserve relationships that survive perspective. It is ruthless about discarding the rest.

• Incidence means a point lies on a line, and that statement remains meaningful under projection.
• Collinearity means points share a line, and lines remain lines after projective maps.
• Concurrency means lines share a point, and that stays true even when the picture looks skewed.

What projective geometry does not promise is the stuff your ruler and protractor want.

• Length and distance
• Angles and perpendicularity
• Midpoints and equal spacing

Right question
When perspective enters, ask which points land on which lines, not how far apart they look.

This is why points at infinity fit naturally. Parallel lines share a direction. Projectively, that shared direction is encoded as a shared intersection point on the line at infinity, so concurrency can talk about parallel families the same way it talks about ordinary intersections.

Homogeneous coordinates make infinity computable

A point in the projective plane is represented by homogeneous coordinates (x:y:z), where scaling does not change the point. (2:4:2) and (1:2:1) name the same point because you can multiply or divide by any nonzero constant.

The Euclidean plane sits inside as the slice z=1. A Cartesian point (X,Y) becomes (X:Y:1). Points with z=0 are the points at infinity. They are not far away. They are directions.

To engage with the conversions and the scaling idea, use this interactive mapping between Cartesian and homogeneous forms.

Lines become equally simple. A line is a triple (a:b:c) meaning the set of points (x:y:z) satisfying the dot product equation $ax+by+cz=0$. Incidence becomes one algebraic check. If a point satisfies the equation, it lies on the line.

Two practical payoffs drop out immediately.

• Intersections are uniform. Parallel lines in Euclidean terms intersect at a point with z=0.
• Computations are stable conceptually. You do not need separate formulas for vertical lines, infinite slopes, or special cases for direction.

Mental model
z=1 is your usual drawing paper, z=0 is the horizon where directions live.

Projection explains why lines stay lines

A central projection takes points in space and maps them to an image plane by drawing a ray from a camera center through the point. The key fact is simple. A 3D line and the camera center determine a plane, and that plane meets the image plane in a line. So straight lines stay straight.

That is the engine behind projective transformations. When you compose projections and changes of coordinates, you get mappings of the form

$$(x:y:z)\mapsto(H\cdot(x,y,z)^T)$$

where H is an invertible 3x3 matrix, and equality is up to scaling because coordinates are homogeneous. This is why projective geometry feels linear once you accept the scaling rule.

Infinity moves because the line at infinity is not sacred under a general projective transform. An affine transform keeps the line at infinity fixed, so parallelism stays parallelism. A full projective transform can send some finite points to infinity and bring some infinite points back to finite ones. That is not a paradox. It is exactly what a change in viewpoint can do.

Cross-ratio is the invariant that survives perspective

The cross-ratio of four collinear points is preserved by every projective transformation, even though distances and midpoints are not.

If $A,B,C,D$ lie on one line, the cross-ratio $(A,B;C,D)$ is a single number computed from directed distances along that line in any affine coordinate. Under a projective transformation, the picture can stretch, compress, and send points flying toward infinity, but the cross-ratio stays the same.

Use the interactive to drag a projective transform and watch which quantities change and which do not.

This invariance is why projective geometry can do measurement indirectly. You may not trust a length in the image, but you can trust a cross-ratio built from points that are easy to detect, like corners and intersections. In computer vision, that is the bridge between a distorted photograph and a stable geometric statement.

Practical tip
If you need one number that perspective cannot fake, reach for cross-ratio.

Duality swaps points and lines

Projective geometry has a symmetry called duality. Many statements remain true if you swap the words point and line, and swap lies on with passes through. It feels like a magic trick until you notice the algebra already treats points and lines similarly. Both are triples, and incidence is one dot product equation.

Try reading these in pairs.

• If two points determine a line, then two lines meet in a point.
• If three points are collinear, then three lines are concurrent.
• If a point lies on a line, then the dual line passes through the dual point.

The interactive comparison helps you practice flipping a statement into its dual and recognizing the same structure underneath.

Duality lets theorems show up in pairs, so proving one can hand you another for free once you translate it correctly. It also trains you to look for structure rather than metric detail, which is exactly the projective mindset.

Degeneracies that break your intuition

Projective geometry removes many Euclidean exceptions, but it introduces its own gotchas. Most are about degeneracy, meaning your configuration loses the generic position assumptions that silent proofs rely on.

A few common ones.

• Coincident points or overlapping lines make intersections ambiguous.
• Three lines that you expected to meet at a unique point may actually share a whole pencil of points if two lines coincide.
• Parallel-as-a-limit cases hide points at infinity that suddenly matter when you switch coordinate charts.
• In computations, nearly parallel lines can produce large coordinate values, and scaling can amplify numeric noise.

Open the reveal panel to see typical failure modes and the simple checks that prevent them from derailing a construction.

When something breaks, look for a hidden limit. Projective geometry often replaces the question why did it fail with which point moved to infinity, or which intersection became non-unique.

Safety check
Before trusting an intersection, ask whether the defining lines are distinct and not numerically almost the same.

Where points at infinity show up

The line at infinity is the horizon in a drawing, but it is also a practical tool in applied geometry.

• In computer vision, vanishing points are points at infinity of world-parallel line families, pulled into the image by projection.
• In perspective drawing and CAD, projective closure turns missing intersections into well-defined ones, so algorithms do not branch on parallel cases.
• With conics, projective ideas explain why ellipses, parabolas, and hyperbolas are the same object viewed with different intersections with the line at infinity.

If you have a real scenario in mind, use the prompt to map it to projective objects, and identify what should be invariant and what is allowed to change.

A good next step is to pick one photograph or sketch you already trust, mark two sets of parallel lines, and locate their vanishing points. You are literally finding points at infinity, made visible by projection.