Digital Jordan Theory

• Digital Jordan Theory is a framework that extends classical Jordan theorems to discrete spaces, defining digital manifolds and robust separation in digital images.
• It employs graph-theoretic models and digital homology to verify topological properties and assess segmentation quality in 2D and higher dimensions.
• Efficient algorithms based on digital Jordan curves facilitate image segmentation, morphing, and compression by ensuring unique interior-exterior separations.

Digital Jordan Theory is the mathematical framework that extends the classical Jordan Curve Theorem—and its higher-dimensional generalizations—to the setting of discrete and digital spaces, particularly focusing on lattices such as $\mathbb{Z}^n$. It provides rigorous definitions, separation theorems, and algorithmic tools to distinguish interiors and exteriors in digital images, segmentations, and more generally in finite or countable grids endowed with graph-theoretical or topological adjacency structures. Central to this theory are concepts of digital manifolds, compatible adjacency relations, and discrete homology, underpinning both foundational theory and applications in digital image analysis, segmentation quality assessment, and image morphing.

1. Classical and Digital Jordan Theorems

The core of Digital Jordan Theory consists of discrete analogues of two foundational results from classical topology:

Jordan Curve Theorem (Classical): Every simple closed curve $C \subset \mathbb{R}^2$ divides the plane into two connected components—an interior and an exterior—with $C$ as their common boundary.

Jordan–Brouwer Separation Theorem (Higher Dimensions): Any locally flat $(n-1)$-sphere embedded in $\mathbb{R}^n$ separates the space into exactly two components, each with the $(n-1)$-sphere as boundary.

In the digital context, these results are recast for subgraphs or point-sets in discrete lattices with suitably chosen adjacency relations. Rosenfeld’s digital Jordan theorem established that, to properly recover separation, one must use a dual pair of adjacencies: a $k$-connected curve ensures the complement splits into two $k'$-connected regions, with $(k, k') = (4, 8)$ or $(8, 4)$ in $C \subset \mathbb{R}^2$0 (Cote et al., 22 Mar 2025).

Generalizations to $C \subset \mathbb{R}^2$1 require careful axiomatic definitions for digital $C \subset \mathbb{R}^2$2-manifolds and “good pairs” of adjacencies: the foreground (the set or surface) and background (its complement) must be assigned adjacencies $C \subset \mathbb{R}^2$3 satisfying separation and double-point exclusion properties (Hünniger, 2011).

2. Graph-Theoretic and Topological Frameworks

Adjacency Graphs: Digital spaces are modeled as graphs, with vertices for points or pixels and edges for adjacency. For 2D images ($C \subset \mathbb{R}^2$4), the principal adjacencies are 4-adjacency (cardinal neighbors) and 8-adjacency (including diagonals). In higher dimensions, generalizations such as “proto-adjacency” $C \subset \mathbb{R}^2$5 and full “$C \subset \mathbb{R}^2$6-adjacency” capture axis-aligned and all immediate neighbors, respectively (Hünniger, 2011, Evako, 2013).

Normal Digital Manifolds: A digital $C \subset \mathbb{R}^2$7-manifold is defined recursively by demanding that every point has a neighborhood (rim) homeomorphic to a digital $C \subset \mathbb{R}^2$8-sphere and that certain local and cube-wise connectivity properties are satisfied. The absence of “double points” (simultaneous crossing of $C \subset \mathbb{R}^2$9- and $C$0-edges) ensures unobstructed separation by digital surfaces (Hünniger, 2011).

Topological Embedding (Khalimsky Plane): The Khalimsky topology on $C$1 and $C$2 equips the grid with a non-Hausdorff Alexandroff topology, allowing concepts such as arcs, curves, and connectedness to parallel the continuum setting. Tools like the slant map and operator $C$3 bridge between the purely combinatorial and the topological formulations, ensuring that theorems in either setting can be systematically transferred and compared (Cote et al., 22 Mar 2025, Kandola, 2019).

3. Digital Jordan–Brouwer Theorem and Manifold Criteria

The digital Jordan–Brouwer theorem asserts: if $C$4 is a digital $C$5-manifold under a good pair of adjacencies $C$6, then the complement $C$7 has exactly two $C$8-connected components and $C$9 is their shared digital boundary (Hünniger, 2011, Evako, 2013). This is verified by associating $(n-1)$0 to a finite simplicial complex $(n-1)$1 encoding all required connectivity and manifold axioms, then applying Alexandrov’s separation theorems for pseudomanifolds in $(n-1)$2.

A succinct criterion is: for a digital curve $(n-1)$3, if $(n-1)$4 is a simple closed $(n-1)$5-curve (every point has exactly two $(n-1)$6-neighbors in $(n-1)$7) and is sufficiently long ($(n-1)$8), then $(n-1)$9 splits into exactly two $\mathbb{R}^n$0-connected components (Cote et al., 22 Mar 2025, Benedictis et al., 15 Jan 2026).

Good Adjacency Pairs: The separation property fails unless the adjacencies inside and outside are chosen asymmetrically. In $\mathbb{R}^n$1D, only $\mathbb{R}^n$2 and $\mathbb{R}^n$3 adjacencies are permitted if one requires separation without ambiguity (Hünniger, 2011).

4. Homology, Betti Numbers, and Segmentation Evaluation

Digital homology offers algebraic invariants to classify surfaces and curves. For subgraphs modeling the boundary $\mathbb{R}^n$4 of a digital shape, Betti numbers $\mathbb{R}^n$5 (number of connected components) and $\mathbb{R}^n$6 (number of independent cycles) can be explicitly computed via graph-theoretic formulas: $\mathbb{R}^n$7

In image segmentation, a binary mask $\mathbb{R}^n$8 is assessed for “Jordan-segmentability” if its extracted $\mathbb{R}^n$9-curve candidate $(n-1)$0 has $(n-1)$1 and if $(n-1)$2 (using 8-adjacency) has exactly two connected components. This unsupervised, topologically grounded criterion is not directly captured by standard pixel-wise metrics (IoU, Dice, precision) and is critical in applications where interior/exterior separation and global shape coherence are essential, such as medical imaging (Benedictis et al., 15 Jan 2026).

5. Algorithmic Extraction and Verification

Concrete extraction and verification of digital Jordan curves proceeds algorithmically on binary images:

1. Preprocess and optionally clean the mask.
2. Extract boundary candidates: find $(n-1)$3-connected foreground pixels $(n-1)$4 that are $(n-1)$5-adjacent to any background.
3. From $(n-1)$6, assemble the boundary $(n-1)$7 by further ensuring $(n-1)$8-adjacency to the background.
4. Build the $(n-1)$9-adjacency graph $k$0 and compute $k$1; enforce single-component, single-loop constraints.
5. On the complement, construct the $k$2-adjacency graph $k$3 and check for exactly two components.

If all criteria are satisfied, the mask is Jordan-segmentable. Counterexamples—such as diagonal slashes or filled disks—demonstrate the necessity and sharpness of these criteria (Benedictis et al., 15 Jan 2026).

6. Digital Jordan Curves in Motion Planning and Complexity

Digital Jordan curves can be considered as combinatorial objects in finite Alexandroff topological spaces (finite $k$4 spaces or “COTS”). The parameter space of such curves on a digital domain $k$5,

$k$6

is path-connected: any Jordan curve can be deformed into any other through a sequence of elementary moves that shrink or expand the interior, or adjust localized curve blocks. The topological complexity $k$7 quantifies the minimal number of continuous “motion planning rules” needed to morph any Jordan curve into another; it is finite, and explicit algorithms exist for constructing such paths (“fences” of curves) (Kandola, 2019).

These concepts underpin practical image morphing, segmentation, and compression algorithms, allowing the continuous transformation of segmented images by topologically valid Jordan curves.

7. Applications, Extensions, and Comparative Perspective

Digital Jordan Theory has pervasive impact in:

• Image segmentation: Certifying that a segmentation partition defines a unique inside/outside structure.
• Medical imaging: Ensuring that boundaries in diagnostic images correspond to genuine, topologically valid enclosed regions.
• Image morphing and compression: Utilizing the controlled deformation of Jordan curves for transformations and compact representation schemes.
• Algorithmic object counting and connectivity testing: Fast, algebraic or combinatorial checks for boundary integrity in digital images and volumetric data.
• Higher-dimensional digital topology: The digital Jordan–Brouwer theorem gives a blueprint for separating digital $k$8-spaces by digital $k$9-spheres, informing algorithmic topology in dimensions $k'$0 (Evako, 2013).

Traditional metrics, based on local overlap or boundary proximity, cannot fully assess such topological integrity. Digital Jordan Theory supplies a mathematically guaranteed, unsupervised, low-cost structural verification, and enables rigorous algorithmic foundations for contemporary image analysis (Benedictis et al., 15 Jan 2026, Kandola, 2019).

Core Topic	Primary Reference	arXiv id
Jordan-segmentable masks, segmentation metrics	"Jordan-Segmentable Masks: A Topology-Aware definition for characterizing Binary Image Segmentation"	(Benedictis et al., 15 Jan 2026)
Graph-theoretic vs. topological digital Jordan	"Bridging Graph-Theoretical and Topological Approaches: Connectivity and Jordan Curves in the Digital Plane"	(Cote et al., 22 Mar 2025)
Digital spheres and separation in Zn	"The Jordan-Brouwer theorem for the digital normal n-space Zn"	(Evako, 2013)
Axiomatic digital manifolds, good pairs	"Digital Manifolds and the Theorem of Jordan-Brouwer"	(Hünniger, 2011)
Digital Jordan curves, topological complexity	"The Topological Complexity of Spaces of Digital Jordan Curves"	(Kandola, 2019)