index

Medial Axis Extraction in O(n)

Around three months ago, I solved a long-standing problem in computational geometry and computer vision:

Extracting the true medial axis from a binary image in linear time per pixel.

No thinning.
No iterative label propagation.
No graph reconstruction passes.

The method uses parallelizable, local operations and produces a pre-pruned, usable medial axis directly.

It took significant effort to get here. This is not a small improvement over existing approaches. It fundamentally changes the cost and structure of skeletonization.

I hesitated to say anything publicly. Once a result like this is known to exist, others may attempt to rediscover it. That risk hasn’t gone away, but neither has the value of moving first.

So I’m stating it clearly:

I’ve implemented it.
I’ve tested it.
It works.

Why this matters

This has immediate implications across multiple domains:

• OCR / glyph extraction

  Direct, high-quality vectorization of shapes from raster images without heavy ML pipelines.

• Computer vision pipelines

  Faster, simpler structural feature extraction with deterministic behavior.

• Medical imaging (including 3D extension)

  Potential for more efficient and precise structural analysis of volumetric data.

• SIGINT( Signal and intelligence ) applications

  Any domain that depends on extracting topology from noisy binary fields.

On disclosure

This is likely dual-use technology. That complicates traditional patent paths and introduces uncertainty around control and ownership depending on jurisdiction.

I am evaluating paths that preserve both the value of the work and my ability to operate freely.

Investment

I am open to serious discussion with investors or partners who understand the scope and implications of this.

Seeking $20M for a minority stake. Final terms depend on structure, alignment, and strategic value.

If you understand what this unlocks, you understand why that valuation exists.

Contact me.