8
$\begingroup$

I'm trying to digitalise the solid lines in the plot below and find the approach here doesn't work for my case.

If the solid lines can be well highlighted into simpler format, I think this ResourceFunction["ExtractPlotImageData"] would be useful.

As I'm very less experienced in image processing, I'm wondering where to start using related techniques to get the two solid lines.

z0/zd vs. lambdaP

$\endgroup$
3
  • $\begingroup$ There's too much going on in your plot, dashes, arrows, labels and all sorts. First crop it, cropped = ImageTake[img, {10, 270}, {55, 352}]; then put this into an image editing program and erase as much as possible of the lines you don't want. $\endgroup$ – flinty May 21 at 10:19
  • $\begingroup$ yes, it could be the start; but I have many of such plots to process so really want something that automates all the processing. $\endgroup$ – sunt05 May 21 at 10:21
  • 2
    $\begingroup$ Sorry, but that's just not going to happen for you. You're going to get very unreliable data from all the markings and lines that throw off the data extraction. Maybe hire some people to manually mask out the lines. $\endgroup$ – flinty May 21 at 10:27
13
$\begingroup$

What flinty says in his comment is correct: you'll have a hard time fully automating this.

However, we can give it a go for an individual plot and then we'll have a better idea of what's required.

ic = SelectComponents[ColorNegate[MorphologicalBinarize[i, .6]], Large]

enter image description here

Here we've managed to extract the main solid lines, but it's a bit messy and this isn't going to generalise well.

We can clean up these components a little bit (here's even more tunable parameters that will make this difficult to generalise):

d = Pruning[
  MorphologicalTransform[Thinning@Pruning[ic, 10], "Bridge"], 10]

and now we split them into two, by subtracting the branch points (where the two overlap) from the image:

i2 = d - MorphologicalTransform[d, {"SkeletonBranchPoints"}]

enter image description here

If you look very closely you can see they no longer connect. We can verify this with:

Colorize[MorphologicalComponents[i2]]

enter image description here

Now that we've gone this far we might as well finish the job.

Let's crop the image and extract the components.

components = MorphologicalComponents[ImageCrop@i2]

(We crop here since we know that the plot goes from 0 to 1 for one of our lines. If that wasn't the case you'd need to figure out how to rescale the data to the correct axes.)

Let's get the individual components out of this Matrix and turn them into black-and-white images where black is the plot:

plots = ColorNegate@Image@# & /@ 
  Table[(components /. {n -> 1, _Integer -> 0}), {n, 
    Union[Flatten@components][[2 ;;]]}]

And now we can use that handy resource function to extract the values:

ListLinePlot[ResourceFunction["ExtractPlotImageData"] /@ plots]

enter image description here

Snatching defeat from the jaws of victory, our resource function automatically scales the y axis. We'll now need to implement our own version that keeps the y scale...

pd = Table[(components /. {n -> 1, _Integer -> 0}), {n, 
   Union[Flatten@components][[2 ;;]]}]

rescaled =
  SubsetMap[Rescale[#, {0, Dimensions[pd][[3]]}] &,
     SubsetMap[Rescale[#, {0, Dimensions[pd][[2]]}] &, 
      Position[Transpose@Reverse@#, 1], {All, 2}], {All, 1}] & /@ pd;

This is a pretty naive version of that ExtractPlotImageData function and (I'm sure) fails many edge cases.

Anyway, now we've got our plot redrawn!

ListLinePlot@rescaled

enter image description here

And even overlay our one and the original:

Overlay[{
  Show[i, ImagePadding -> {{0, 0}, {0, 6}}],
  ListLinePlot[rescaled, AspectRatio -> .92, 
   ImagePadding -> {{50, 8}, {0, 0}}]
  }, PlotRangeClipping -> False]

enter image description here

I hope it's clear that this kind of process is quite hard to automate, although with a few hours of work you might come up with a Manipulate that does much of what you need with some manual intervention (picking the scale of the lines, picking the threshold for the binarize, and so on).

$\endgroup$
2
  • 1
    $\begingroup$ fantastic! learnt a lot from your post! many thanks! $\endgroup$ – sunt05 May 21 at 13:01
  • 1
    $\begingroup$ Happy to help, best of luck generalising it. $\endgroup$ – Carl Lange May 21 at 13:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.