Retrieving the individual Image Channel Statistics

[yazinsai]

Hi, I'm developing a website that uploads an image and reads the colorspace and Mean for each of the R,G and B channels.
I've been able to read the colorspace using identify -format %[colorspace] rose.jpg

But I'm stuck in trying to figure out how to read the Mean for each Channel..

The output for:

Code: Select all

identify -verbose rose.jpg

includes the details i'm looking for:

Code: Select all

...
  Channel statistics:
    Red:
      min: 0 (0)
      max: 255 (1)
      mean: 231.673 (0.908523) [b]<!-- This is what i'm after[/b]
      standard deviation: 66.5237 (0.260877)
      kurtosis: 7.3051
      skewness: -3.00254
    Green:
      min: 0 (0)
      max: 255 (1)
      mean: 227.871 (0.89361) [b]<!-- This too[/b]
      standard deviation: 70.4308 (0.276199)
      kurtosis: 5.16135
      skewness: -2.59435
    Blue:
      min: 0 (0)
      max: 255 (1)
      mean: 224.366 (0.879867) [b]<!-- And this, please[/b]
      standard deviation: 77.2012 (0.30275)
      kurtosis: 3.677
      skewness: -2.35433
...

What is the most efficient way of extracting this information from the image?

[anthony]

The "sed" command can search for and extract the desired part of the line.

Code: Select all

identify -verbose rose: | \
   sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p'

Red
0.57142
Green
0.35004
Blue
0.315562
Overall
0.412341

[yazinsai]

Works! Thanks Anthony

[anthony]

Ask no questions, and you shall be told no lies

Comment on your signature...
Ask no questions and the world will not advance!

But this quote comes to mind...

All men must die, it was their single common heritage.
But a book need never die and should not be killed;
books were the immortal part of man.
-- Robert A. Heinlein, "Farnham's Freehold"

[yazinsai]

Hey is there any way that I can get the colorspace in the same command so it's more efficient than running "identify" twice? Thanks!

P.S: Nice quote Mine's not fitting for a forum where questions are meant to be asked. Got it changed

The "sed" command can search for and extract the desired part of the line.

Code: Select all

identify -verbose rose: | \
   sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p'

Red
0.57142
Green
0.35004
Blue
0.315562
Overall
0.412341

[anthony]

Hey is there any way that I can get the colorspace in the same command so it's more efficient than running "identify" twice? Thanks!

Code: Select all

identify -verbose rose: | \
   sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p; /Colorspace:/p'

You can either find and modify the line using s/..../...../p or just print the line using /..../p

"sed" is a very versitile 'stream editor', and you can write some very complex text processing with it. Others that are in some ways simplier, but with a different focus, is "awk" and "perl".

Note however that this is 'shell' specific, and not really ImageMagick, but I help when I can.

[fmw42]

why not just do:

convert rose: -format "%[colorspace]\n%[fx:mean.r]\n%[fx:mean.g]\n%[fx:mean.b]\n%[fx:mean]" info:

sRGB
0.57142
0.35004
0.315563
0.412341


P.S. when I do

identify -verbose rose: | \
sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p'

I do not get the names: red, green, blue and overall listed to the terminal as Anthony showed earlier!

[yazinsai]

You Sir are amazing! I didn't find the mean.r/g/b references in the -format page! THANK YOU!!

why not just do:

convert rose: -format "%[colorspace]\n%[fx:mean.r]\n%[fx:mean.g]\n%[fx:mean.b]\n%[fx:mean]" info:

sRGB
0.57142
0.35004
0.315563
0.412341


P.S. when I do

identify -verbose rose: | \
sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p'

I do not get the names: red, green, blue and overall listed to the terminal as Anthony showed earlier!

[yazinsai]

Thanks Anthony! I did find a guide that explains how SED works in detail, but need to get around to reading that sometime. Really appreciate your help..

Hey is there any way that I can get the colorspace in the same command so it's more efficient than running "identify" twice? Thanks!

Code: Select all

identify -verbose rose: | \
   sed -n 's/.*\(Red\|Green\|Blue\|Overall\):.*/\1 /p; s/.*mean:.*(\(.*\)).*/\1/p; /Colorspace:/p'

You can either find and modify the line using s/..../...../p or just print the line using /..../p

"sed" is a very versitile 'stream editor', and you can write some very complex text processing with it. Others that are in some ways simplier, but with a different focus, is "awk" and "perl".

Note however that this is 'shell' specific, and not really ImageMagick, but I help when I can.

[fmw42]

You Sir are amazing! I didn't find the mean.r/g/b references in the -format page! THANK YOU!!

You have to do that via fx escapes. see
http://www.imagemagick.org/Usage/transform/#fx_escapes
and
http://www.imagemagick.org/script/fx.php

[yazinsai]

I knew I probably missed it; thanks again!

You Sir are amazing! I didn't find the mean.r/g/b references in the -format page! THANK YOU!!

You have to do that via fx escapes. see
http://www.imagemagick.org/Usage/transform/#fx_escapes
and
http://www.imagemagick.org/script/fx.php