Shape Elongation via Isotropic Skeletonization

18 concepts3 papers1 proposal

Human vision does not measure length like a physical ruler. If it did, an object's perceived length would remain constant regardless of its thickness, contrast, or orientation. Yet, psychophysical evidence shows that our perception of geometry is highly malleable: wider objects often look shorter than narrow objects of the exact same physical length, and faint contours can appear shorter than bold ones.

To explain these anomalies, early computational vision researchers proposed that the brain computes length through a dynamic pipeline of geometric reduction. Instead of measuring raw boundaries, the visual system transforms shapes into simplified internal "skeletons" to judge elongation.

The Two-Stage Reduction Pipeline

How does the brain transform a complex, two-dimensional shape into a singular judgment of length? In Conjectures on the Perception of Elongation, Edward S. Deutsch outlines a two-stage image-processing mechanism that models this perceptual feat:

  1. Convex Hull Formation: The visual system first constructs a convex hull—a minimal bounding envelope—around the stimulus. This step abstracts away internal gaps, line drawings, or fragmentations, treating the enclosed area as a unified, solid geometric volume.
  2. Isotropic Thinning: The system then subjects this hull to a uniform reduction process, eroding the boundaries of the shape inward at an equal rate from all directions. As the shape shrinks, it collapses into a core medial axis, or a "skeleton."

The core insight, also detailed in the parallel paper Conjectures, is that our perception of elongation is directly tied to the length of this internal skeleton. If the uniform reduction causes a shape to collapse entirely into a single point or a non-elongated cluster, the image is not perceived as elongated at all.

Why Width Compresses Perceived Length

This skeletonization model provides an elegant, intuitive explanation for why an object's aspect ratio distorts its perceived length. Because the thinning process is isotropic—meaning it occurs at a uniform rate from all directions—the width of an object directly impacts how its longitudinal skeleton is formed.

Imagine a wide rectangle versus a narrow line of the same length. For the wide rectangle, the isotropic erosion must operate longer to collapse the shape to its center line. However, because the erosion happens uniformly from all directions, it does not just eat away at the long sides; it also erodes the shape from the short ends at the exact same rate.

Wide Object:  [  |  ]  --> More thinning needed --> Ends erode inward further --> Shorter Skeleton
Narrow Object: [-----]  --> Less thinning needed --> Ends preserved           --> Longer Skeleton

Consequently, the extensive thinning required for wider shapes prematurely clips the ends of the internal skeleton, resulting in a compressed core line. A narrow object requires very little thinning to find its center, preserving the full length of its skeleton. This is why wider geometric hulls inherently feel shorter.

Coupling Intensity with Spatial Extent

The geometric reduction model explains how shape boundaries interact, but how does the strength or clarity of those boundaries affect our perception? In Human Vision Paradox Implicates Relaxation Model, Fanya S. Montalvo addresses a complementary phenomenon: the perceived geometric dimensions of a contour are fundamentally influenced by its visual intensity or distinctness.

When a contour is faint or weakened—whether due to low contrast or neural adaptation—it is perceived as physically shorter. Montalvo explains this by framing vision as a "relaxation model" governed by local neural constraints rather than independent feature detectors.

If the visual system conflates the magnitude of a neural response (signal intensity) with spatial extent, a lower-intensity signal effectively shrinks the perceived boundary. In the context of skeletonization, a less distinct boundary provides a weaker constraint during the reduction process, causing the internal representation of the contour to recede or shorten.

Maintaining Perceptual Stability

Why would the human visual system rely on a reduction process that introduces these systematic distortions? The answer lies in the need for orientation invariance.

As noted by Deutsch, an object's perceived elongation must remain stable whether it is oriented horizontally, vertically, or diagonally. By relying on a uniform, omnidirectional thinning process applied to a simplified convex envelope, the visual system ensures that its structural analysis of an object is invariant to rotation. The visual system accepts minor distortions in length and width as a computational trade-off to achieve a stable, orientation-independent understanding of shape structure.

Key Foundational Work

Conjectures on the Perception of Elongation (1971) — contributed 8 concept elements to this direction.
Conjectures on the Perception of Elongation (1971) — contributed 6 concept elements to this direction.
Human Vision Paradox Implicates Relaxation Model (1977) — contributed 4 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_contour_distinctness_bias_b4354db3,
  title = {contour distinctness bias},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/contour-distinctness-bias/rdr_b4354db38dca482a9f770edea952ac5e},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_contour_distinctness_bias_b4354db3_v1,
  title = {contour distinctness bias},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/contour-distinctness-bias/rdr_b4354db38dca482a9f770edea952ac5e/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake