4

Is it possible to get the predominant colour in an image from the command line? Ideally I'd like it back as hex or RGB.

I thought it might be possible to do this with imagemagick's identify command, but I couldn't see an option for doing this in the documentation.

2
  • Could you be more specific about what is a predominant color for you? 24-bit RGB pixel has 16777216 possible color values and I doubt if finding the most repeating one is useful in any way.
    – gronostaj
    Apr 2, 2013 at 16:54
  • I'd like to narrow it down to a fairly small number of colours... I'm no expert on colours, but the sixteen HTML safe colours or the X11 colour names would be ideal. I guess this would need some kind of process like (i) reduce the colour space of the image (ii) create a table of pixels by colour (iii) map the dominant colour to the closest colour name. No idea how to do any of that, though!
    – Richard
    Apr 2, 2013 at 16:58

2 Answers 2

13

Imagemagick's identify -verbose provides a palette histogram table, but only if the number of unique colors is below a limit (1024 in the versions I've checked, v6.x up to v6.8.3).

Instead you can just "convert" your image to a histogram, this output is not limited by number of unique colors:

convert image.jpg  -format %c -depth 8  histogram:info:histogram_image.txt
sort -n histogram_image.txt | tail -1

Sorting the output numerically by the first column sorts by frequency of pixel colors. (Frequency of specific colour pixels might not correspond to a human perception of predominant color of course.)

If you're reducing colors, you can probably just trust convert to do it for you by your choice of dithering and/or posterizing. Even specifically to the web-safe 216:

convert image.png  +dither -remap netscape:  image_websafe.png

You would probably get closer to a perceived predominant color by some combination of blurring, resizing and posterizing, this is not a simple problem to define and solve:


See also the following link for dcolors, a script which uses Imagemagick to determine a set of predominant colors in an image, uses include coordinating your desktop color scheme with your chosen desktop image: http://javier.io/blog/en/2015/09/30/using-imagemagick-and-kmeans-to-find-dominant-colors-in-images.html

5
  • 1
    Thanks - great answer and the first two commands actually work pretty well! One final question: what do you think is the best way to map the predominant hex value to the nearest websafe value? Is this just a question of hex arithmetic or is there a more sophisticated way?
    – Richard
    Apr 2, 2013 at 21:35
  • Given the palette is so limited, treating (R,G,B) as a 3d space and finding the minimal Euclidean distance to a defined colour would probably suffice for most purposes. You can get convert to do this for you by adding -remap netscape: after -depth 8 in the first command above. Apr 2, 2013 at 22:02
  • it might work better to convert to the HSV colour space, then only consider the "H" portion in the histogram. Is this possible with IM?
    – Michael
    Jul 12, 2020 at 22:46
  • You can add "-colorspace HSV" before -format in the first example to show HSV instead of sRGB in the histogram output, but you will need to process the hsv() values separately I think. If you use -modulate 0,0,100 to perform a transform in HSL or HSV, you get a zero hue too, per-spec. Perhaps something smarter using -fx or -color-matrix could work, but that's beyond me today... Jul 13, 2020 at 16:26
  • on windows, this should be run through Cygwin
    – ivan866
    Oct 22, 2020 at 21:23
7

I found this elegant and short answer here http://blog.endpoint.com/2011/04/determining-dominant-image-color.html

$ convert Waffle.jpg -scale 1x1\! -format '%[pixel:u]' info:-
rgb(219,166,94)
2
  • It doesn't work any more
    – Buzut
    Feb 12, 2019 at 21:15
  • I think I properly updated the command to work with magick instead of the original convert command.
    – StarGeek
    Jul 17, 2023 at 19:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .