Limiting or Minimizing Histogram Output

[dmfelder]

I need to determine the count of "approximately-white" pixels in my art. I can write histogram output, import it & analyze, but the file is so large.

I use:

Code: Select all

convert input.jpg -format %c histogram:info:out.txt

and I get a very large list like this:

• 396: (129,129,130,255) #818182 rgba(129,129,130,1)
  573: (129,130,130,255) #818282 rgba(129,130,130,1)
  294: (130,129,129,255) #828181 rgba(130,129,129,1)
  16: (130,130,129,255) #828281 rgba(130,130,129,1)
  2170: (130,130,130,255) #828282 grey51
  707: (130,130,131,255) #828283 rgba(130,130,131,1)
  1032: (130,131,131,255) #828383 rgba(130,131,131,1)
  1: (130,202,160,255) #82CAA0 rgba(130,202,160,1)
  203: (131,130,130,255) #838282 rgba(131,130,130,1)
  4: (131,131,130,255) #838382 rgba(131,131,130,1)

It's hugely redundant data (you don't need the second set of RGB values). If I could NOT print the second set of values, my file would be 1/2 the size. In fact, I only need the RGB OR the HEX (not both). And to top it off, I really only want to see "near-white" pixel quantities (north of 253, 253, 253).

I have tried preceding the code with the command:

Code: Select all

-define histogram:unique-colors=false

but I still don't get what I want.

All I need is the count and EITHER the RGB or the HEX.

Any other ideas?

Thanks so much!

[snibgo]

For starters, I would pre-process the image to turn black all the pixels that were red<253 OR green<253 OR blue<253, or whatever conditions you want. This may eliminate most lines on the listing.

If you just want a count of the number of pixels that are approximately white, then turn them all exactly white and the others exactly black. Then your histogram has only two colours. You don't even need a histogram: "identify -verbose" would give the answer.

[fmw42]

try

Code: Select all

convert input.jpg -threshold XX% -format %c histogram:info:out.txt

where XX% is very large, as close to 100% as you want your tolerance for white.

[dmfelder]

Thanks to both of you. Basically, Snibgo's idea is realized in Fred's threshold command.

This is a very simple and effective solution. I think the threshold command will create a maximum of 3 lines of output:

• Closer to Black: 63: ( 0, 0, 0,255) #000000 black
  Closer to White: 25: (255,255,255,255) #FFFFFF white
  Transparent: 12: (255,255,255, 0) #FFFFFF00 rgba(255,255,255,0)

Given that my file has now reduced to a max of 3 lines, I have a follow-up question...

Is there a way to capture the variables instead of writing the output to a file? I tried using Identify, but I didn't get any result. Ideally, I'd love to avoid writing the file in the first place. If there is anyway to populate a variable with the data that is being written, I could avoid the complication of write/read.

Thanks

P.S. For reference, I use commands like this to obtain image data:

Code: Select all

imagedata = ObjMi.Identify("-units", "PixelsPerInch", "-format", "%x %y %w %h", myimagename)

[fmw42]

On Unix, this will put the data from the histogram into a variable. From there you can use sed and cut to get the parts you want as new variables.

Code: Select all

data=`convert rose: -threshold 95% -format "%c" histogram:info:`
echo "$data"
      3080: (  0,  0,  0) #000000 black
       140: (255,255,255) #FFFFFF white

[snibgo]

P.S. For reference, I use commands like this to obtain image data:

Fine, but you don't say what language that is. I probably wouldn't know the answer for your language, even if you did say. You should look at documentation for that language.