Using on Windows about doxa HOT 5 CLOSED

You just need to install R (at least version >4) at https://cran.r-project.org/bin/windows/base/

Then in R install the package as follows:

install.packages("image.binarization")
install.packages("magick")

And run the code where you replace scan.jpg with the path to your image.

library(magick)
library(image.binarization)
img <- image_read("scan.jpg")
img <- image_convert(img, format = "PGM", colorspace = "Gray")
img
img_blackwhite <- image_binarization(img, type = "su")
img_blackwhite

img_blackwhite<- image_binarization(img, type = "sauvola", opts = list(window = 75L, k = 0.2)
img_blackwhite
image_write(img_blackwhite, "test.png")

from doxa.

I'm on Windows and wrote this R wrapper at https://cran.r-project.org/web/packages/image.binarization/index.html.
Examples at https://cran.r-project.org/web/packages/image.binarization/readme/README.html and more examples by looking at the documenation of ?image_binarization

from doxa.

Thank you so much, @jwijffels

I sheepishly admit -- I actually downloaded image.binarization yesterday to see if I could figure it out...and I couldn't. :(

I have a feeling that at least in part it has to do with "the package requires a compiler with C++17 capabilities." I don't even know if I have this, much less how to do it. (Ha.)

For what it's worth, I understand a very little of Python, use Visual Studio Code, and my goal is for image results more than programming. (I'm here because I'm looking for an optimal binarization method for separating sometimes faded glyphs from backgrounds (with multivariate tones from aging and the like) on a historical manuscript. I came across Doxa and image.binarization searching for "improved Sauvola" in Google.)

from doxa.

Very kind of you to walk me through this. Success! I have it working in R now on Windows. Very much appreciated.

from doxa.

Good to know it worked - I'm using the R wrapper for image cleaning in order to find text lines in scans of 18th-19th century documents https://github.com/DIGI-VUB/image.textlinedetector as well as some preprocessing in Handwritten text recognition at https://github.com/DIGI-VUB/HTR-tests.
Great toolkit, Doxa!

from doxa.