It’s a deeper issue than image processing though. This is just an allegory (and I’m proud to FINALLY have figured out a good process for turning bmp into vector)

It’s a deeper issue than image processing though. This is just an allegory (and I’m proud to FINALLY have figured out a good process for turning bmp into vector)

Does the data drive the thresholding though?
I don’t think it does. A decent algorithm / NN process can do a good autothresholding but the judgement call of utility is still set or approved by human.

—-

You’re learning while teaching.
That’s what good teachers do. The most crucial network forms inbetween your network and the network of the database. It’s a dance, a relationship. More like gentle coaxing than demanding.

====

This thing I just did today ties in. Started out trying to make a Bmp into a Vector by hand but ended up stumbling into a deeper issue: setting a threshold. Doorway between rooms. Categories. State Rights vs Individual Rights. Take your pick of threshold, it’s always for us to set and decide.

====

There are a lot of them. But I don’t know of any that’s universal. I’m hoping to come across a list of them with code so I can compare the thought processes behind them.

ImageJ has an auto-thresholding that has a lot of options and names the algorithms. That’s how I got the montage. (very hard to find a program that will just give me a threshhold montage but ImageJ (or Fiji, which is ImageJ with all the good plugins) has it.

=====

 

Medical imaging has been dealing with this issue for decades now: “How to extract volumetric medical data from a 2D image?”
 
 
====
Rotislav is smart. That’s a MapReduce method (or Newton’s squares I forget). The shrinking takes care of the worst of the gaps and brings pixels together
 
Funny thing is I knew that and yet have been taking the long road anyway. Thank you – I’d forgotten!
====

Leave a comment

Your email address will not be published. Required fields are marked *


7 − = six

Leave a Reply