Jump to content

 
Photo

Online tool : Visualization of color proportions within an image

- - - - -

  • Please log in to reply
3 replies to this topic

#1
l.jegou

l.jegou

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 112 posts
  • Gender:Male
  • Location:Toulouse
  • Interests:Webmapping, thematical mapping.
  • France

Hi, for my PhD i'm looking at the map image as a whole, and developing some analysis tools.

One of them is an online tool to represent the colour proportions of an image (pixels relative quantities), using frequencies with a HSV (hue / saturation / brightness value) colour model:

http://www.geotests....rammes2-en.html

This representation aims to simplify the perception of hues and their relative proportions in a cartographic image. This can be useful to examine the results of colour choices for the legend of a map, especially gradients. (But any image can be used).

The tool is completely online, it's HTML5+Javascript, so very open-source :-) (No upload required, your browser is working locally).

I'm hoping for comments, thanks :)

#2
l.jegou

l.jegou

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 112 posts
  • Gender:Male
  • Location:Toulouse
  • Interests:Webmapping, thematical mapping.
  • France

Anyone ?

#3
P.Raposo

P.Raposo

    Contributor

  • Validated Member
  • PipPip
  • 47 posts
  • Gender:Male
  • Location:Penn State
  • United States

Hi l.jecou,

That seems like a really neat tool, nice bit of programming :) I tried it with several images I uploaded to it. As requested, comments:

It took me a minute to figure out that the graph was a plot of HSV values - hues at angles, saturation at radii lengths, and values in concentric circles. Perhaps a very brief primer, or tooltip, for the user to explain this, something easier to get to than your "legend" link?

If I understand the symbology correctly, I'm worried about the concentric circles for value part. Doesn't the area (size) of the circle describe the frequency of pixels at the given hue/saturation? If so, doesn't stacking values make it possible to occlude smaller circles (of lesser value) below? I think you're handling that by making sure larger circles are lower and smaller are higher in the stack, but it's possible two or more circles can be so close in size, or equivalent, that you wouldn't get to see the "rings" they define.

You could make the program plot a 3D HSV model, and give people the ability to manipulate it (say, Processing, or a Java Applet...?)

Many maps are vector formats. If the tool could take vectors (e.g., SVG, or PDF from ArcMap, or Esri .mpk map packages), and analyze the output of the drawing, that would be cool too :)

Also, minor detail: the "Nb. of main colors to exclude" field was cut off in my layout (I'm using Opera 11.64); all else was fine.

Cheers, nice work, merci, P

#4
l.jegou

l.jegou

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 112 posts
  • Gender:Male
  • Location:Toulouse
  • Interests:Webmapping, thematical mapping.
  • France

Thanks, P. Raposo for your comments. Some answers :

The link labeled "legend" is perhaps not enough visible, i'll modify it.
You wrote about uploading files, but as i mentioned in my presentation, no upload is done, the application is working locally on your machine (albeit in a browser), that's the benefit of HTML5.

The concentric circles are indeed stacking, so their surfaces are not fully visible. It's a recurrent problem in thematic cartography, but as i was interested in colors, i couldn't use the usual solutions (transparency, hollow circles). I think that the interactivity can help to help about that : one can change the parameters to view the masked circles. And there is the 3D interactive model, where no symbol is masked because you can always pivot the representation (but it's not as simple to use or to find, it seems you have not tried it.)

I really have a user interface / help problem here :/

The vector format is really dominant in published cartography (as a pro print cartographer, i can say so, too :)), but it's not very practical to use for color analysis. The rasterisation of vector maps are very easy to do, though.

Merci for taking time to test and comment, it's very much appreciated :-)




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->