Is the apparent contour the same as the contour generator?

No. The generator lies on the three-dimensional surface; the apparent contour is its projection into the image plane.

Does every contour belong to the outer outline?

No. Self-occlusion can produce visible contours inside the overall silhouette, and visibility conventions matter.

Can one outline recover an object's complete 3D shape?

Generally not. Many shapes can produce the same projection, and hidden or inward-facing geometry may remain unconstrained.

AI Glossary / Computer vision and multimodal AI / Advanced

Apparent contour

Definition

An apparent contour is the image-plane projection of a surface's contour generator: locations where the viewing direction lies in the surface's tangent plane. In vision and graphics, its visible parts describe important occluding boundaries, including parts of an object's outline.

What it means

A smooth object can have a clear outline even when no physical crease exists there. A ball's visible edge is an example. The surface continues around the ball, but the viewing rays graze it at the outline instead of meeting it more directly.

The corresponding three-dimensional surface locations form a contour generator. Their projection produces the apparent contour. Visibility must also be considered: some projected contour portions can be hidden behind other parts of the surface or other objects. Terminology varies: the cited Algebraic Smooth Occluding Contours paper reserves apparent contour for the visible projection; this entry uses the broader projection convention and states visibility separately.

The outer silhouette boundary is especially familiar, but an object can also have self-occluding contours inside that outer boundary. A folded or curved shape may hide part of itself, creating an internal visible contour without a change in surface material.

How it works

For a smooth surface point, a contour-generator condition is that the surface normal is perpendicular to the viewing direction. Under orthographic projection, that direction is constant. Under perspective projection, it depends on the camera position and the surface point.

A rendering algorithm can locate these surface points or approximate them on a mesh, project them into the image and determine which portions are visible. In an actual photograph, a vision algorithm must infer suitable boundaries from image evidence rather than reading the surface geometry directly.

Not every detected image edge is an apparent contour. Paint boundaries, shadows and highlights can create strong image gradients on a continuous visible surface. Distinguishing these sources of edges is essential when contours are used to infer shape or constrain reconstruction.

Three practical examples

A line-drawing renderer extracts visible occluding contours from a 3D model to communicate its shape with a small number of lines.
A multi-view reconstruction method uses silhouette and occluding-boundary information to constrain possible surface geometry.
A product-inspection system compares an object's visible outline with an expected projection under a controlled camera setup.

Think of it this way

Think of looking at a smooth hill: the skyline marks where the ground turns away from your view, even though there is no painted line or sharp ridge at every point on that skyline. The analogy is limited because real skylines may include physical corners, vegetation and multiple overlapping objects, whereas the smooth-surface definition uses a precise tangency condition.

Worked example

Consider a sphere of radius R centered at the origin, viewed orthographically along the z direction. Its surface satisfies x² + y² + z² = R², and its normal points in the radial direction (x, y, z).

Perpendicularity to the viewing direction requires z = 0. The contour generator is therefore the circle on the sphere satisfying x² + y² = R² and z = 0. Projecting onto the image's x-y plane gives the familiar circular outline.

For R = 2, the projected contour satisfies x² + y² = 4. The sphere has no physical seam along that circle. Moving the viewing direction changes which surface points form the contour generator, even though the sphere itself has not changed.

Illustrative case study

Fictional example: Formline Tools compares rendered views of a smooth handle design. Designers use contour drawings to see where the silhouette narrows and where one curved section hides another.

They notice that a dark stripe in a shaded rendering does not appear in the geometric contour drawing. Inspection shows that the stripe is a material marking, not an occluding boundary. The team keeps those cues separate when reviewing the shape.

For a later camera-based inspection trial, the engineers control viewpoint and lighting and compare visible outlines with expected projections. This fictional case illustrates the distinction between surface geometry and image appearance, not a claim that contours alone determine every dimension of the handle.

Common misconceptions and mistakes

An apparent contour is not necessarily a physical edge or a permanent curve painted on the object. Its generating surface points can change as the viewpoint changes.

A silhouette also does not reveal every concavity or hidden surface. Different 3D shapes can share the same outline from one view. Treating every high-contrast line as an occluding contour can mistake shadows or texture for geometry. The smooth tangency condition needs special care at corners, boundaries and singularities, where a simple surface normal may not provide the full description.

Business use

Contour information is valuable when shape matters and texture is unreliable or unavailable. It can support visual design review, model rendering and controlled inspection, while providing constraints for reconstruction.

Define whether a system uses the outer silhouette, internal occluding contours or all geometric contour candidates. Control or estimate the camera geometry, and assess visibility rather than assuming every projected curve is observable. In measurement workflows, compare multiple views or additional cues when a single outline leaves important shape ambiguities unresolved.

Check your understanding

A smooth ball has a circular outline in an image. Must there be a physical circular seam on the ball at that location?

Answer: No. The outline arises from viewing rays tangent to the smooth surface. The relevant surface points depend on the viewpoint.

Related terms

Orthographic projection
Contour generator
occluding contour
silhouette
surface normal
tangent plane
visibility
shape from silhouette

Learn next

Rotate a smooth 3D object and compare its visible contour with a fixed painted line on its surface. Observe which points move along the contour as the view changes. Then add a shadow and a material stripe, and practice separating those image edges from boundaries caused by occlusion.

Sources

Shan and colleagues, Occluding Contours for Multi-View Stereo, author project: https://grail.cs.washington.edu/projects/sq_rome_g2/ Capouellez and colleagues, Algebraic Smooth Occluding Contours, research paper: https://arxiv.org/abs/2306.01973 The sphere calculation and design case are original illustrations. Geometric contour candidates and their visible subsets are distinguished explicitly.

Previous: Anytime algorithm Next: Appearance model