Matt Zucker, a robotics and computer vision engineer, writes about image processing, graphics and math projects, with code and visual examples.
HN commenters link his Noteshrink post as a well-written, HN-worthy piece on compressing scanned notes, and also point to his map vectorization work.
This is fascinating, there is quite a bit of overlap with a font that I started designing recently: http://alanbernstein.net/images/line-font-lowercase
I do small projects on CNC machines, and I got annoyed at having to trace both sides of a character, when using the outline of a standard font. In response, I created the single-stroke characters shown above, but only enough of them to draw basic labels.
Part of my original inspiration was this blog post: https://mzucker.github.io/2016/08/03/miniray.html
It would be interesting to do a closer comparison of these two fonts, it looks like some of our letters are identical. I definitely prefer some of Leon's characters, especially the 's'. It should be easy enough for me to implement mine in javascript, because I focused on simplicity of the glyph definitions. For example, lowercase 'e' is defined like this:
'e': [Line([-0.5, 0.5], [0.5, 0.5]), Arc([0, 0.5], 0.5, 0, 7/4)],
I hope I can find the time to experiment with this.
Looks great. I'm not the target audience, but feature 5 caught my attention
>We will soon be releasing an open source library for users to generate HTML from Latex Flavored Markdown
For some reason, the app reminded me about enhancing scanned copy of hand-written notes and I managed to search it online [1] I couldn't remember where I had saved it in my bookmarks, after finding it online, I found the HN entry as well [2]
[1] https://mzucker.github.io/2016/09/20/noteshrink.html
[2] https://news.ycombinator.com/item?id=16567275
Following the handwritten comments brought me there, which I found to be HN-worthy: https://mzucker.github.io/2016/09/20/noteshrink.html
Also, as a sign of our times, it hyperlinks to what might have been a Microsoft tech page/knowledge base in the past, now completely replaced by poor taste LLM advertising slop. Good work, Microsoft.
> Are there tools that can detect the spine in the middle, de-skew the image, adjust discolouration, detect text, and OCR?
https://www.flameeyes.eu/projects/unpaper provides a partial solution. https://mzucker.github.io/2016/09/20/noteshrink.html for some more ideas
This is fantastic, I would only suggest one more thing, yesterday there was article on first page about vectorization, so further compression and cleaning could be done if text would convert to vectors, links is:
https://mzucker.github.io/2018/05/14/maptrace.html
http://kellysutton.com
Stumbled upon this one yesterday, it's from a paper, but well written:
https://mzucker.github.io/2016/09/20/noteshrink.html
A long time favorite writer:
http://www.frankchimero.com/writing/the-webs-grain/