Legacy ImageMagick Discussions Archive

Hello, all.

First of all, I want to make it clear that I have tried searching for this not only on Google but on these forums, but I haven't been able to find a solution. I apologize if this topic has already been brought up.

Essentially, I want to take an image and shift the RGB values of every pixel within that image. For example, I am recoloring a bright blue image. I want to shift the RGB values of every pixel by the following amounts in order to achieve a brownish image:

R: -74
G: -109
B: -142

A pixel with a RGB value of (128,160,190) becomes (54,51,58), for example.

Is there any way to do this using ImageMagick's command-line interface? My only option right now seems to be to manually change the values in GIMP which has proven to be a painstakingly slow and arduous process.

Hi ...
I suppose that using the "-color-matrix" option in convert would be your answer to your problem...
As the web usage docs referer it it could e a way to solve your problem ...

Something in like :
convert -color-matrix " 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0" imagein imageout

Although you would need to calculate the "correct values" for the matrix in order to achieve what you need ...

One possibility is to use "-evaluate". But I think, you have to adjust the subtraction values to your QuantumRange here (e.g. by multiplying them with 257 for a Q16 version). Example:
convert input -channel R -evaluate subtract 19018 -channel G -evaluate subtract 28013 -channel B -evaluate subtract 36494 output

-color-matrix has offsets

see

http://www.imagemagick.org/script/comma ... lor-matrix

also see http://www.adobe.com/devnet/flash/artic ... ns_04.html

but set the offsets to fractions (-79/255=-.310, -109/255=-.427, -142/255=-.557)

convert rose: -color-matrix \
"1 0 0 0, 0, -0.310 \
0 1 0 0, 0, -0.427 \
0 0 1 0, 0, -0.557 \
0 0 0 1, 0, 0 \
0 0 0 0, 1, 0 \
0 0 0 0, 0, 1" \
rose_adjusted.png

Drarakel wrote:One possibility is to use "-evaluate". But I think, you have to adjust the subtraction values to your QuantumRange here (e.g. by multiplying them with 257 for a Q16 version). Example:
convert input -channel R -evaluate subtract 19018 -channel G -evaluate subtract 28013 -channel B -evaluate subtract 36494 output

This solution produced an outcome closest to what I was trying to achieve. I compared the ImageMagick-produced image side-by-side with an image that I recolored manually in GIMP, and the RGB values were almost 100% spot-on.

When I tried to produce the same result with -color-matrix, the saturation and hue were off. The results were not accurate enough.

Thank you to all that helped me!

It is likely I made an error in my calculations of the fractions as the following all give exactly the same results:



convert rose: -channel R -evaluate subtract 19018 \
-channel G -evaluate subtract 28013 \
-channel B -evaluate subtract 36494 \
rose_evaluate1.png


rval=`convert xc: -format "%[fx:74*100/255]" info:`
gval=`convert xc: -format "%[fx:109*100/255]" info:`
bval=`convert xc: -format "%[fx:142*100/255]" info:`
convert rose: -channel R -evaluate subtract $rval% \
-channel G -evaluate subtract $gval% \
-channel B -evaluate subtract $bval% \
rose_evaluate2.png


rval=`convert xc: -format "%[fx:-74/255]" info:`
gval=`convert xc: -format "%[fx:-109/255]" info:`
bval=`convert xc: -format "%[fx:-142/255]" info:`
convert rose: -color-matrix \
"1 0 0 0, 0, $rval \
0 1 0 0, 0, $gval \
0 0 1 0, 0, $bval \
0 0 0 1, 0, 0 \
0 0 0 0, 1, 0 \
0 0 0 0, 0, 1" \
rose_colormatrix.png

compare -metric rmse rose_evaluate1.png rose_colormatrix.png null:
0 (0)

compare -metric rmse rose_evaluate2.png rose_colormatrix.png null:
0 (0)

Well, that's equally interesting. Thank you for that.

I think I'm going to stick with the "-evaluate subtract" method, though, as it involves far less typing and leaves far less room for my n00biness to make errors.

If you are going to resort to FX why not just subtract the color!
NOTE -evaulate uses Actual values are not normalized values for subtract (-function uses normalized)

Assuming a 8 bit IM

convert rose: -channel R -evaluate subtract 74 \
-channel G -evaluate subtract 109 \
-channel B -evaluate subtract 142 \

For IM Q16 you will need to use a value of value*65535/255

Of course use a percentage value then it is value*100/255 just as Fred specified above for ALL versions of IM.