Add channels to CMYK file output

[ahhhaweb]

I need to output CMYK image files with two additional channels for alpha(white) and varnish. I understand that internally IM can have extra channels. I'm looking for a way to write those channels to a TIFF output file. Is this something that might be in the works or is there a way to accomplish this that I haven't found?

[snibgo]

As far as I know, IM cannot (yet) write extra channels ("metachannels") to TIFF, or any format except "ftxt:".

EDIT: Correction: metachannels are also written to the "mpc:" format.

[ahhhaweb]

Thanks for the quick response Snibgo and for confirming what I had been told.

If anyone has any suggestions for alternatives that can be done in Ubuntu webserver environment I would appreciate hearing ideas.

[fmw42]

Can you not just keep individual images for the extra channels and combine them with the main image as needed?

[urban-warrior]

ImageMagick 7 was designed to handle meta-channels internally and thus far only the FTXT and MPC image formats fully support the extra channels. We need to increase our support, so let's start with your use case. Attach a Zip archive to this post with your original source image and any extra channels with a note on how you would like the TIFF image to handle the extra channels. We currently support CMYKA so its likely just a question of supporting the extra channel you call varnish.

[fmw42]

Can you post your Photoshop file showing the desired result? I cannot say that I really understand your description as is.

[ahhhaweb]

Yes, I will try to have a sample file posted tomorrow to provide an example of a properly working six channel file.

[urban-warrior]

We have two images, one sRGB and one Gray+Alpha. We want to convert these 5 channels to a 6 channel TIFF in the CMYA colorspace with an extra channel populated from the Gray image. Let us know if this is not what you are looking for. Also, we will of course need a few days (or weeks) to complete the work and post our progress here.

[ahhhaweb]

No, thanks for asking for clarification urban-warrior. A use case as you describe is probably the more common scenario but in our case we just want the image data from the grayscale threshold(3) image replicated to all 6 channels in a tiff file. I only included the rgb jpg because it was the source file that the grayscale was derived from but other than that the rgb jpg is irrelevant to the task. I'll try to get you a sample 6 channel grayscale tiff later today.

[ahhhaweb]

Here is an example of the 6 channel tiff file that we want to be able to generate from IM.
threshold_6 channel.zip

[fmw42]

How about a syntax such as:

magick grayimage.suffix -duplicate 5 -define tiff:extra-channel=1 -set colorspace CMYKA image.tiff

[urban-warrior]

The latest commits to the main repo support multispectral images. This means

magick threshold_6_channel.tif  new.tif
identify -verbose threshold_6_channel.tif  new.tif

properly clones the source image. We have generalized support to other combinations of channels and image formats, however, we will evolve the support as to reduce risk in breaking existing workflows. Instead, we need additional use cases, one at a time, until we have a set of multi-spectral images and use cases we can use for testing any generalized multispectral solution.

[ahhhaweb]

A review of this documentation on the channels feature from Adobe PS seems to confirm the above described use case. It appears that the primary requirement when adding a channel is " All new channels have the same dimensions and number of pixels as the original image."
https://helpx.adobe.com/photoshop/using/channel-basics.html

[urban-warrior]

We now have support for multispectral images in the PSD, TIFF, MIFF, MPC, and FTXT formats. We could benefit from use cases to validate our patches. Post multispectral images in the PSD or TIFF format here and we will validate them against the latest ImageMagick release.

[ahhhaweb]

What a blessing to have folks like you and the IM team contributing such immense amounts of talent and time to build tools that do so much to help the work of others!! I can't tell you how much I appreciate everything you folks do!!

[ahhhaweb]

I feel like a real dummy here coming back to this so many months later but we're having trouble understanding the correct command syntax to generate the 6 channel tif. We're starting with an rgb png file. (sample file attached) We want to clone the image data in that file to all 6 channels in the tiff output file.

[ahhhaweb]

To be clear, we want to
copy or duplicate the data from the source image into all 6 channels in the tiff output file.

[ahhhaweb]

Here is an example of what the resulting output file should contain
layer3 (1).zip

[urban-warrior]

@snibgo, we'll need a patch for this to work with -channel-fx. In the mean-time, we suspect it can be done with the FTXT format. Can you provide a workflow?

[fmw42]

I tried playing with -channel-fx and could not get it to work saving to TIFF.

[snibgo]

The input file 219971487-b828a14f-a545-4133-a7c1-3cb2091b985c.png shown above is 1050x746 pixels, with RGBA channels. Alpha is entirely opaque. The three colour channels are equal.

We want to clone the image data in that file to all 6 channels in the tiff output file.
To be clear, we want to copy or duplicate the data from the source image into all 6 channels in the tiff output file.

Here is an example of what the resulting output file should contain: layer3 (1).zip

The file layer3 (1).tif contains two images. I assume the output TIFF should have only one image. According to IM -verbose both images are CMYK, with an alpha channel, and one meta channel, for a total of six channels. The CMY and meta channels appear to be the same as each other, and alpha is different.

I assume we want to copy the input RGB channels to an output CMYKA plus one meta channel.

Write pixels from the input file (Windows syntax):

magick ^
  219971487-b828a14f-a545-4133-a7c1-3cb2091b985c.png ^
  -channel RGB ^
  -define ftxt:format=\x,\y:\c,\c\n ^
  ftxt:21997.txt

The first three lines from 21997.txt are:

0,0:65535,65535,65535,65535,65535,65535
1,0:65535,65535,65535,65535,65535,65535
2,0:65535,65535,65535,65535,65535,65535

As required, we have six values per pixel.

Read those into a new image, and save as a tiff file:

%IMG71062%magick ^
  -size 1050x746 ^
  -colorspace CMYK ^
  -define ftxt:format=\x,\y:\c\n ^
  -define ftxt:hasalpha=true ^
  -define ftxt:nummeta=1 ^
  ftxt:21997.txt ^
  x.tiff

exiftool x.tiff | grep -i Sample

exiftool says:

Bits Per Sample                 : 16 16 16 16 16 16
Samples Per Pixel               : 6
Extra Samples                   : Unknown (2 0)

So that looks good.

The magick commands can be piped, so the intermediate text data need not be saved in a file.

The input is sRGB but the output is CMYK with the same pixel values, so colours will be reversed when viewed.

[danielljunior]

Hi there! I happen to be the programmer who is working for @ahhhaweb and we're trying to pull this off.
Thanks for your answer @snibgo, I tryied it but I got an error.

PS G:> magick .\Cat_layer5.png -channel RGB -define ftxt:format=\x,\y:\c,\c,\n ftxt:21997.txt
PS G:> magick -size 2800x2200 -colorspace CMYK -define ftxt:format=\x,\y:\c\n -define ftxt:hasalpha=true -define ftxt:nummeta=1 ftxt:21997.txt x.tiff
magick.exe: NumChannelsError `21997.txt' @ error/ftxt.c/ReadFTXTImage/731.

I'm using windows power shell (heretic, I know), magick version 7.1.0-62.

The actual goal is to take an image, no matter the channel format and then take a given channel of said image and copy to all 6 channels of a tiff image.
In our use case, the image is rgb but as you can see as in the examples given by @ahhhaweb it's always grayscale images. The reason they are rgb is because they are generated on a canvas in the browser and that's just how they come, therefore in our use case no matter which of the rgb channels we copy to all 6 channels of the dest image since they (rgb) are all the same.

I'm new to image magick, so I apologize if I say something stupid here. Based on what I read on the documentaion, the -channel-fx seems to be what make sense to use in that regard, but I'm still trying to wrap my head around all the options.

[snibgo]

For writing the ftxt file, I wrote:

-define ftxt:format=\x,\y:\c,\c\n

You have changed this to:

-define ftxt:format=\x,\y:\c,\c,\n

... with an extra comma. When the ftxt coder reads the text file, the number of commas tells it how many channel values are on that line. An extra comma implies an extra channel, hence NumChannelsError.

The actual goal is to take an image, no matter the channel format and then take a given channel of said image and copy to all 6 channels of a tiff image.

You can use -channel to write only that channel to the ftxt file. But you want to repeat the value six times, so repeat \c six times, with a comma between each (but no trailing comma):

magick in.png -channel R -define ftxt:format=\x,\y:\c,\c,\c,\c,\c,\c\n ftxt:out.txt

[urban-warrior]

@snibgo, what if we want CMYK and two meta channels, no alpha? We tried:

magick -size 1050x746 -colorspace CMYK -define ftxt:format="\x,\y:\c\n" -define ftxt:nummeta=2 ftxt:21997.txt x.tiff

and received a CMYK TIFF image with 4 channels. We were expecting CYKK with 2 extra "unspecified" samples. Is the issue with the TIFF writer? Or with FTXT? Or with internal channel handling inside ImageMagick?

[snibgo]

... what if we want CMYK and two meta channels, no alpha?

From my reading of the TIFF standard, this should be possible. Of the TIFF "extra samples", only one channel should be either "1 = Associated alpha data (with pre-multiplied color)" or "2 = Unassociated alpha data", but any number may be "0 = Unspecified data". [EDIT: and there may be no alpha channels.]

I have tried this, and couldn't make it work from IM. The ftxt writer and reader seem to be okay. You can test this by inserting "+write ftxt:test.tmp" before writing the TIFF file, and you will get six values.

The issue seems to be in the TIFF writer in coders\tiff.c. I don't entirely follow the processing, but it seems that if the image has no alpha channel, then tiff.c will not write an alpha to the TIFF (which is fair enough) but nor will it write any meta channels to the TIFF. So the TIFF will have no extra samples at all.

I think this is a bug in coders\tiff.c.

[urban-warrior]

Thanks for the analysis. We'll apply a patch this weekend.

[ahhhaweb]

Thank you all for your support in this quest!