stippler 0.2.1.1

Hand-drawn weighted Voronoi stippling pipeline (Python 3.9.10 / Rhino CPython compatible)

Rhino Grasshopper-Compatible Stippling Processor

for parameters used here, see Example Parameters

---

This is a replication of the following article:

Weighted Voronoi Stippling, Adrian Secord. In: Proceedings of the 2nd International Symposium on Non-photorealistic Animation and Rendering. NPAR ’02. ACM, 2002, pp. 37– 43.

where the author introduced a techniques for generating stipple drawings from grayscale images using weighted centroidal Voronoi diagrams as in the traditional artistic technique of stippling that places small dots of ink onto paper such that their density give the impression of tone.

Authors and credits

• Original replication — Nicolas P. Rougier (BSD license, 2017), replicating Adrian Secord, Weighted Voronoi Stippling, NPAR 2002.
• Python 3.9.10 port, revised hand-drawn stippling pipeline, Lu et al. scientific-illustration extensions, and Grasshopper / Rhino integration — Max Benjamin Eschenbach.

Pre-requisites

The original replication was written and tested on OSX 10.12 (Sierra) using Python 3.6, Numpy 1.12, Scipy 0.18, Matplotlib 2.0 and tqdm 4.10.

This revision ports and extends that code for Python 3.9.10, matching the Rhino 8 / Grasshopper CPython runtime. The port replaces the removed scipy.misc.imread with a small Pillow-based loader; the compute core remains plain numpy / scipy.spatial. It has been verified with:

• Python 3.9.10
• Numpy 2.0
• Scipy 1.13
• Pillow 11.3
• Matplotlib 3.9
• tqdm 4.x

Create the environment with conda:

conda create -n stippler -c conda-forge python=3.9.10 numpy scipy matplotlib pillow tqdm

Original data is in the data directory and you can also obtain it from Adrian Secord homepage.

Installation

Install from PyPI:

pip install stippler

For local development from a clone:

pip install -e .

This installs the dependencies (numpy, scipy, Pillow, tqdm, matplotlib), the importable package stippler, and a stippler console command.

Usage (classic stippler)

The original replication CLI is preserved as the stippler.classic module (run it with python -m stippler.classic ...):

 usage: python -m stippler.classic
                    [--n_iter n] [--n_point n] [--save] [--force]
                    [--pointsize min,max] [--figsize w,h]
                    [--display] [--interactive] file

 Weighted Vororonoi Stippler

 positional arguments:
   file                  Density image filename

 optional arguments:
   -h, --help            show this help message and exit
   --n_iter n            Maximum number of iterations
   --n_point n           Number of points
   --pointsize (min,max) (min,max)
                         Point mix/max size for final display
   --figsize w,h         Figure size
   --force               Force recomputation
   --save                Save computed points
   --display             Display final result
   --interactive         Display intermediate results (slower)

Hand-drawn pipeline (stippler command)

The stippler console command (module stippler.pipeline) wraps the relaxation in a grayscale-image → stipple-output pipeline and adds three controls that make the result read as hand drawn rather than machine generated. The compute core depends only on numpy, scipy and Pillow (Rhino-friendly); matplotlib is used only for preview rendering.

1. Early-stopped relaxation — fewer Lloyd iterations means the points never settle into the regular hexagonal lattice, so spacing stays organically uneven. This is the single biggest dial.
   • --n_iter n — hard cap on iterations (lower = looser).
   • --epsilon d — optional: stop when mean point movement drops below d.

To control contrast (denser blacks / cleaner whites), use --gamma g. It applies a power curve d ** g to the density that drives point placement, relaxation and dot size, so g > 1 (e.g. 2.0–2.5) concentrates the same n_point dots into the dark areas and thins the light ones; g < 1 flattens the tonal range; g = 1 is the original linear mapping. (--threshold is a blunter, hard cutoff that forces lighter greys to pure white.) 2. Varying dot size — a min/max radius spread plus per-dot random jitter breaks the constant-radius "machine stipple" giveaway.

• --pointsize min max — radius range (density drives the base size).
• --size_jitter f — per-dot multiplicative radius noise, e.g. 0.15.

3. Imperfect placement & edges — Gaussian positional jitter plus wobbly, non-circular dot outlines.
   • --position_jitter s — Gaussian noise std on final positions.
   • --edge_noise f — dot-edge wobble as a fraction of radius, e.g. 0.08.
   • --edge_segments n — vertices per dot outline.

Output format follows the --out extension (.png, .pdf, .svg).

stippler data/boots.jpg --n_point 12000 --n_iter 8 \
         --pointsize 0.8 3.0 --size_jitter 0.2 --position_jitter 0.5 \
         --edge_noise 0.09 --seed 3 --out data/boots-stipple.png

The pipeline can also be imported as a library: stippler.stipple(...) returns a StippleResult carrying points, radii and per-dot polygons (handy for feeding geometry into Rhino later), which render_matplotlib, render_svg and save_points consume.

Scientific-illustration pipeline (Lu et al.)

The hand-drawn pipeline is further extended with optional controls adapted from Lu et al., Non-Photorealistic Volume Rendering Using Stippling Techniques (resources/vis_stipple.pdf). Each feature is independently toggled via IllustrationParams (library) or the CLI illustration argument group (stippler --help):

• Boundary — boost stipple density on high-gradient edges.
• Silhouette density — concentrate dots on view-facing silhouette regions (uses an optional normal map).
• Interior — sparse stipples in flat, low-gradient areas.
• Lighting — modulate density from inferred or supplied normals.
• Depth attenuation — thin stipples in far regions (requires a depth map).
• Gradient size — scale dot radius by local gradient magnitude.
• Silhouette curves — extract and draw feature-line strokes over the dots.

Normal and depth maps should be aligned with the beauty-pass image. In Rhino, the Grasshopper component can capture these automatically when the options that need them are enabled.

Grasshopper / Rhino integration

Max Benjamin Eschenbach integrated the revised pipeline into Grasshopper for Rhino 8 CPython 3.9.10:

• grasshopper_userobjects/STIPPLER_StippledViewCapture.ghuser — drop-in user object that captures the active viewport and writes a _stippled output (PNG, SVG, or PDF), including hand-drawn controls and the Lu et al. options above.
• grasshopper_userobjects_src/ — Python 3 script source and input/output reference. Re-export the user object from Grasshopper after editing the script (right-click → Save User Object…).

Install the stippler package into the Rhino CPython environment the component uses (see Installation above), then paste or sync the script into a GH Python 3 component.

Example Parameters

Parameters used for the Donut example

Parameter	Value
Resolution	1920 x 1080 px
NumDots	100000
Iterations	12
Gamma	1.0
RMin	1.0
RMax	6.0
SizeJitter	0.15
PositionJitter	0.1
EdgeNoise	0.1
Seed	42
DPI	300
Boundary	false
SilhouetteDensity	false
Interior	false
Lighting	false
DepthAttenuation	false
GradientSize	false
SilhouetteCurves	true
CurveWidth	0.5