AI Glossary / Computer vision and multimodal AI / Beginner

Parallel projection

Definition

Parallel projection maps three-dimensional points onto a two-dimensional plane along parallel projection lines. Unlike perspective projection, its scale does not shrink merely because an object moves farther from the viewing plane. Orthographic projection is a common special case.

What it means

A perspective photograph makes a distant object appear smaller than an equally sized nearby object. Parallel projection removes that particular depth cue. This is useful when a viewer needs a consistent geometric representation rather than a camera-like impression. Orthographic projection uses lines perpendicular to the image plane; oblique projection uses parallel lines at an angle. Both belong to the broader parallel-projection family. Parallel lines generally remain parallel after projection, although a line pointing along the projection direction can collapse to a point. The result is still two-dimensional, so some depth information is necessarily lost.

How it works

Choose a viewing direction and a plane, then project every scene point along that same direction. In a simple orthographic view looking along the z-axis, the point (x, y, z) becomes (x, y), followed by any chosen display scaling and translation. The z-coordinate can still help determine which surface hides another, even though it does not set image scale.
Rotating the scene before projection produces different orthographic views. An isometric view is a particular orientation often used to show several sides of an object. Oblique projection instead lets depth contribute a fixed linear displacement on the image plane. Neither construction uses the division by depth characteristic of ordinary perspective projection.

Three practical examples

• A product engineer inspects the front and side views of a component with consistent drawing scales.
• A scientific visualization tool uses an orthographic camera to compare the visible sizes of objects at different depths.
• An illustrator uses an isometric view to present the layout of equipment without perspective convergence.

Think of it this way

Think of a set of perfectly parallel guide rails carrying every point toward a drawing board. The rails do not fan out from a single observer. This explains the geometry, but real light and ordinary cameras do not generally behave like this ideal construction. The analogy also does not mean that every three-dimensional length is preserved on the board.

Worked example

Consider two equal horizontal bars, each 20 millimeters long, parallel to the image plane. One is at depth 100 millimeters and the other at depth 200 millimeters. An orthographic display scale of 3 pixels per millimeter makes both bars 60 pixels long. Moving either bar farther away does not change that projected length.
Now turn one bar so that it points partly into the depth direction. Its projected length can become shorter because of orientation. Equal depth-independent scale therefore does not mean that arbitrary lengths can be measured directly from one view. The distinction is between depth-related perspective shrinking and orientation-related foreshortening.

Illustrative case study

In a fictional furniture studio, reviewers compare handles on a cabinet using a rendered perspective view. The nearer handle appears larger, causing a discussion about an apparent design mismatch. A designer switches to a front orthographic view and checks the model dimensions. The handles have equal dimensions; their differing apparent sizes came from perspective. The studio keeps the perspective rendering for the customer presentation and includes orthographic drawings for dimension review. This illustrative case shows why the projection should match the question, without suggesting that an attractive rendering proves manufacturing accuracy.

Common misconceptions and mistakes

Parallel projection is not identical to every possible orthographic view: orthographic is one subset, and oblique views also use parallel projection lines. It does not preserve all distances or angles in three dimensions. A view can hide important features and overlapping surfaces. Changing the camera projection also does not correct inaccurate model geometry. Finally, a normal photograph should not be treated as an orthographic measurement simply because the subject looks flat. Perspective effects, lens distortion and scale calibration may still matter.

Business use

For product communication, simulation and technical review, the main benefit is a predictable display of geometry. Teams should specify the view orientation, display scale and purpose of each export. When measurement matters, use the actual model coordinates or an appropriate calibrated workflow instead of estimating hidden dimensions from a screenshot. Evaluate whether users need visual realism, geometric comparison or both. Providing separate views can prevent a single image from being asked to serve incompatible purposes.

People also ask

Is parallel projection the same as orthographic projection?

Orthographic projection is a form of parallel projection with perpendicular projection lines. Oblique parallel projection is another form.

Do distant objects become smaller?

Their scale does not change merely because of depth, although orientation and clipping can affect what is visible.

Can it show depth?

Shading, overlap and a rotated view can suggest depth, but the projection itself discards one spatial degree of freedom.

Check your understanding

Two equal bars parallel to the image plane appear equally long in an orthographic view. Does this prove that every tilted bar of the same true length will also look equally long?

Answer: No. Tilt can shorten a bar's projection even though distance from the image plane does not create perspective shrinking.

Related terms

• Orthographic projection
• Perspective projection
• Image plane
• Foreshortening
• oblique projection
• isometric view
• camera model
• occlusion

Learn next

Learn perspective projection next and compare the role of depth in its equations. Then rotate a simple box under orthographic and perspective cameras, observing which size relationships change and which features become hidden.

• Perspective projection

More in Computer vision and multimodal AI

• Oriented projective geometry - Oriented projective geometry is a form of projective geometry that retains a distinction between opposite homogeneous directions.
• P-frame - A P-frame, or predictive frame, is a coded video picture that can use previously decoded reference pictures to predict image content.
• Panorama - A panorama is an image showing an extended view of a scene, often created by combining overlapping photographs into one continuous composition.
• Path finding - Path finding is the process of finding a sequence of valid connections from a starting state to a destination.
• Pattern grammar - A pattern grammar is a set of rules describing how simpler components can combine into structured patterns.
• PDE / diffusion inpainting (Bertalmío) - PDE-based inpainting fills a selected missing or damaged image region using equations that propagate information from its surroundings.

Sources

VTK, vtkCamera class reference, including parallel-projection controls: https://vtk.org/doc/nightly/html/classvtkCamera.html VTK Textbook, camera and projection discussion: https://raw.githubusercontent.com/lorensen/VTKExamples/master/src/VTKBookLaTeX/VTKTextBook.pdf Primary-source search excerpts consulted; full source and editorial review remain pending.

Previous: Parallel programming model Next: Parallel tempering