Beefy Boxes and Bandwidth Generously Provided by pair Networks
Problems? Is your data what you think it is?
 
PerlMonks  

Re: pixel counting on a fouling panel

by Arguile (Hermit)
on Jun 09, 2003 at 01:09 UTC ( #264199=note: print w/ replies, xml ) Need Help??


in reply to pixel counting on a fouling panel
in thread pixel counting

This is untested; just going from memory and a quick look at the Image::Magick API Reference. Basically it should read in the image, figure out which colour is transparent, then loop through every pixel and increment a counter when the pixel isn’t transparent. Again, it’s untested. I’m unsure if the pixel index starts at 1 (as I used) or 0. I’m also not totally sure on the $image->Get('transparent') call -- read the API for the exact method -- it’s been a while since I've used it. This should give you a good jumping off point though.

use Image::Magick; my $image = new Image::Magick; $image->Read('foo.png'); my $count; # The number of non-tra +nsparent pixels. my $transparent = $image->Get('transparent'); # Index of the transpar +ent colour. my $height = $image->Get('height'); # Image height. my $width = $image->Get('width'); # Image width. for my $x ( 1 .. $height ) { for my $y ( 1 .. $width ) { $count++ if $image->Get("pixel[$x,$y]") != $transparent; } }

If I get what you're trying to do, it might be beneficial to store each species as a mask. If you save it as a black and white bitmap counting becomes really fast, since you can read the file as one long binary string and just count the 1’s.

Another possibility (again I’m assuming you're generating these masks by hand) is to use a format that allows multiple layers. Storing the image as a photoshop PSD or something similar would allow you to keep all the masks as named channels or layers. The Gimp might be a great tool here as it allows Perl extensions and has said layers. This way the masks stay embedded right in the original image for easy archiving, and you could even store the result of those counts in the image comments. With a little Perl magic, Gimp could become a biology workdesk :).


Comment on Re: pixel counting on a fouling panel
Select or Download Code
Re: Re: pixel counting on a fouling panel
by redbeard (Beadle) on Jun 09, 2003 at 03:40 UTC
    Fantastic - thanks! Unfortunately, I can follow neither of your two above suggestions, although don't think I haven't contemplated it. Unfortunately a variety of the species involved have multiple color morphs, or are exactly the same color as another species (I mean you have to zoom in on individual zooids in a colony to tell the difference- hence we're using a high-res digital underwater camera).

    Saving different layers in a PSD would be nice, but the problem there is that often we have species overgrowing each other in a non-lethal way (e.g. a colonial sea squirt overgrows a mussel. Both live, and both need to be counted). The current scheme I'm going with is copying the area of each species, putting it into its own .gif, and using some filename information for the final data file.

    Although if you know a way to do all of that in GIMP, I'd love to hear it!

      As requested in CB, some clarification...

      From what you’ve said you’ll be hand drawing outlines of what you believe to be the extent of each species boundry; even if another species happens to be overgrowing that boundry to some extent. In this case you’ll be guessing the exact extent and copying that whole area into a new image.

      What I propose is creating a layer or channel** for each species in the image. So for each species:

      1. Create a layer or channel
      2. Turn on their layer (to around 50% transparency so you can see the image underneath)
      3. Use a paintbrush to mask the areas you think they exist in, even if partially covered by another species.
      4. Turn off the layer.

      This will essentially give you a one bit mask that represents the living area of each species. The layers can all overlap without interfering with each other. In addition all the data is stored in one file, easy to keep track of or change.

      This could also work very well for quick visual surveys if you wanted to get fancy. Assign each species-layer a colour and then turn the ones of interest on with a high transparency. You’ll see a montage of interacting colour representative of that species living areas and their overlaps.

      Anyways, enough talk. On to some code!

      use Gimp qw(:auto); use Gimp::Fu; # ... my @species = qw( mussels squid octocerata ); # Our species of intere +st. # Create a layer for each species (labelled for that species) and add +it to # the image. foreach my $specie ( @species ) { my $layer = gimp_layer_new( $img, $w, $h, RGB, $specie, 50, NORMAL +_MODE); gimp_image_add_layer($layer, -1); } # ...

      Adding a zillion layers (one for each specie) to each image would be boring manual work (colouring will be boring enough), so in the above partial snippet we let a Gimp script do that for us. In production you’d probably pull this from a file or database. Anything else you'd normally do to preprocess the image could be done here as well, for example autolevels or auto contrast.

      It’s late, so I think I’ll stop here. For now I’ll leave you with some reference material to look over.

      See Also:


      ** Layers and channels are different, which you use will depend on exactly what you want to do with it. If all you need is a one bit mask then channels would be your best choice. For simplicity sake I’ll refer to the choice as layers throughout the post.

      *** Spiffy title eh? :)

        Fantastic, and thanks so much for the reference material! THis method may well work, esp if I can then incorporate a way to count pixels per layer into the script and then dump the output into a file (read in the current species mentioned, add any new species from this particular image, reoutput the old data with 0's in corresponding new species, then output a line of the new data).

        Just gotta brush up on my Gimp-perl!

        Although I may still use the cut-into-seperate-images method, as I'm thinking of spreading this script around to some other folk I know who do similar analyses, and they're not so gimp-savy, although they're all OSX users.

        If only I could make this a photoshop plugin as well...

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://264199]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others taking refuge in the Monastery: (6)
As of 2014-08-02 04:29 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    Who would be the most fun to work for?















    Results (54 votes), past polls