Efficient Fractal Boundaries via Hierarchical Jittered Voronoi

This article describes a method for generating fractal cell boundaries by recursively nesting Jittered Voronoi grids. By linking sites in fine grids to parent sites in coarser grids, the author creates a hierarchical partition that can be traced back to a root site. The technique is computationally efficient, allows for per-point calculation, and includes an adaptive optimization for faster rendering.

Key Points

• Fractal Voronoi partitions are created by recursively nesting jittered grids and linking sites to their nearest neighbors in the layer above.
• The final partition of any point is defined by the root site (Layer 0) reached when tracing the lineage of the nearest site at the deepest layer.
• The algorithm is computationally efficient because it can be calculated per-point without constructing the entire diagram, using hashed PRNGs for site generation.
• Practical implementation involves either a fixed recursion depth or an adaptive early stopping condition based on an 'influence radius'.
• The resulting boundaries mimic the self-similar, fractal nature of real-world geographic features like coastlines.

Sentiment

The overall sentiment is positive and aligned with the article. Commenters mostly accept the technique as useful and visually convincing, especially for procedural maps, while the limited critical discussion focuses on practical implementation caveats rather than disputing the article's approach.

In Agreement

• The generated regions resemble natural and administrative borders, making the technique attractive for procedural maps and games.
• The method's tendency to produce enclaves and exclaves is viewed as a useful emergent property rather than a flaw.
• A reader reports applying the idea to an existing Voronoi-based Scotland clan map and finding the visual result improved.
• Commenters appreciate the article's clear, concise, and accessible explanation of the technique.

Opposed

• There is no direct rejection of the article's main point, but one commenter warns that related feature-placement use cases can produce repetitive fine-scale artifacts if domain warping is not scaled with cell size.
• A follow-up asks for clarification about what kinds of features the practical caveats apply to, indicating that some implementation guidance may need more concrete examples.