Post

Morphology Dilation and Erosion

Morphology Dilation and Erosion

🧠 Core idea
Dilation and erosion are not filters in the convolution sense.
They are set-based geometric operations defined by how a shape
(the structuring element) interacts with another shape (the image).

This post explains dilation and erosion mathematically,
and then reconnects the math back to intuition.


🔢 Two Viewpoints in Morphology

Morphological operations can be defined in two equivalent ways:

  1. Set-theoretic (binary images)
  2. Lattice / min–max (grayscale images)

Both describe the same idea:
👉 how shapes occupy space.


🟦 Binary Morphology (Set-Theoretic Definition)

Let:

  • Set of foreground pixels (the image) \(A \subset \mathbb{Z}^2\)

  • Structuring element \(B \subset \mathbb{Z}^2\)

  • Translation \(B_x = \{ b + x \mid b \in B \}\)


➕ Dilation (Binary)

Definition

\[A \oplus B = \{ x \mid (\hat{B})_x \cap A \neq \varnothing \}\]

where:

  • Reflected structuring element:
\[\hat{B}\]

Reflection of (B) about the origin.


Interpretation

A point (x) belongs to the dilated set if:

any part of the structuring element,
when placed at (x), overlaps the object.

So dilation answers:

“Does the object touch the structuring element here?”


➖ Erosion (Binary)

Definition

\[A \ominus B = \{ x \mid B_x \subseteq A \}\]

Interpretation

A point (x) remains after erosion if:

all of the structuring element,
when placed at (x), fits inside the object.

So erosion asks:

“Does the object fully contain the structuring element here?”


🟨 Grayscale Morphology (Min–Max Form)

For grayscale images:

  • ( f(x) ) = image intensity
  • ( b(y) ) = structuring element (flat or weighted)

➕ Dilation (Grayscale)

\[(f \oplus b)(x) = \max_{y \in B} \big[ f(x - y) + b(y) \big]\]

If the structuring element is flat:

  • ( b(y) = 0 )

then:

\[(f \oplus B)(x) = \max_{y \in B} f(x - y)\]

Meaning

Dilation selects the maximum value in the neighborhood defined by (B).

This:

  • expands bright regions
  • fills small dark gaps
  • thickens structures

➖ Erosion (Grayscale)

\[(f \ominus b)(x) = \min_{y \in B} \big[ f(x + y) - b(y) \big]\]

For a flat structuring element:

\[(f \ominus B)(x) = \min_{y \in B} f(x + y)\]

Meaning

Erosion selects the minimum value in the neighborhood.

This:

  • shrinks bright regions
  • removes small bright noise
  • thins structures

🧠 Duality Between Dilation and Erosion

Dilation and erosion are duals:

\[(A \oplus B)^c = A^c \ominus \hat{B}\]

This means:

  • one can be expressed in terms of the other
  • they form the fundamental building blocks of morphology

🔍 Why This Is Not Convolution

ConvolutionMorphology
Weighted sumMax / Min
LinearNonlinear
Signal-basedShape-based
Frequency interpretationSpatial occupancy interpretation

Morphology replaces averaging with extreme selection.


💡 Computer Vision Takeaway

Dilation asks “does anything fit?”
Erosion asks “does everything fit?”

That single distinction explains:

  • shape growth vs shrinkage
  • noise behavior
  • why opening and closing work

✨ Summary

  • 🔹 Dilation and erosion are defined by set inclusion
  • 🔹 Binary morphology uses overlap and containment
  • 🔹 Grayscale morphology uses max–min operations
  • 🔹 They are nonlinear but local
  • 🔹 All morphological operators are built from these two

🧱 Once you understand dilation and erosion, morphology becomes intuitive.

This post is licensed under CC BY 4.0 by the author.