Hi @cerbul13,

You'll need to give more information: what OS are you using, what libvips interface are you using, what command did you run, what images did you use as input, what was the error, what result did you expect, etc.

PANTONE make you pay to turn names into numbers, so libvips can't implement anything like that, unfortunately. libvips usually represents spot colours as an array of decimal densities, but the details depend on the libvips interface you are using (CLI, python, ruby, php, etc.).

Hello,
I am using 8.17.4 version(as of changelog), under windows. My goal is to recompose the cmyk image from separations in order to make sure that the final result is similar to what is available in the pdf itself that was processed to obtain the separations.

The command lines I am using:

1. vips.exe bandjoin "CMYK_1(Cyan).tiff CMYK_1(Magenta).tiff CMYK_1(Yellow).tiff CMYK_1(Black).tiff CMYK_1(PANTONE\ 376\ C).tiff" "D:\PDFRIP\Work\CMYKTempComp_1.tiff"[compression=lzw]
2. vips.exe copy "D:\PDFRIP\Work\CMYKTempComp_1.tiff" "D:\PDFRIP\Work\CMYKTempInv_1.tiff"[compression=lzw] --interpretation cmyk
3. vips.exe invert "D:\PDFRIP\Work\CMYKTempInv_1.tiff" "D:\PDFRIP\Work\CMYKComposite_1.tiff"[compression=lzw]

If I combine Cyan, Magenta, Yellow, Black only, without "Pantone 376 C", it will yield correct results, without any Pantone obviously.
If I try to combine all 5, it will interpret Pantone 376 C as if it would be a black background with white pixels. (as opposed to what it was initially, white background with black pixels, as if it would invert the pantone image while using white color as pantone information). So the conclusion is that it would not draw the Pantone image based on "Pantone 376 C" name, instead it interprets this as if the pantone color would be white, and the background of the image as black. (instead of green pixels on white background).
Please let me know if you need files or pictures of the pdf/result/expected result.

For me is also ok to have my files renamed until the operation is done from "CMYK_1(PANTONE\ 376\ C).tiff" to "CMYK_1(CMYK: 10,20,30,40).tiff" or maybe "CMYK_1(Pantone\ 376\ C:[CMYK:10,20,30,40]).tiff". It ok either way as long as I can pass the exact RGB/CMYK numbers to the command line. Please help me with this. I can also prepare a list of pantone conversion manually in my software, if needed.

Ah so these are spot colours and your TIFF is a monochrome halftone? Yes, libvips can apply the CMYK to the mono image, but it's out of scope for the command-line interface, you'd need to use something like python for that.

I could help write the code (just a few lines?), if that'd be useful.

I made a test image like this, hopefully it's similar to your TIFFs:

$ vips extract_band ../theo.jpg mono.tif 1
$ vips relational_const mono.tif "spot (CMYK:10,20,30,40).tif[bitdepth=1,compression=ccittfax4]" moreeq 128

That makes a file called "spot (CMYK:10,20,30,40).tif" containing a 1 bit compressed TIFF, with 1 for where the green band in the JPEG is more than 128. It's not halftoned, but it should be OK for testing.

Here's the scrap of code you need:

#!/usr/bin/env python3

import sys
import re
import pyvips

filename = sys.argv[1]

# parse spot colour from filename
match = re.match(r'.*\(.*CMYK:(\d+),(\d+),(\d+),(\d+).*\).*', filename)
if not match:
    print(f"CMYK specification not found in {filename}")
    sys.exit(1)
spot = [int(match.group(i)) for i in range(1, 5)]

# load the image, make a solid block of the spot colour, mask with the pattern
# in the image, save again
image = pyvips.Image.new_from_file(filename, access="sequential")
cmyk = image.new_from_image(spot).copy(interpretation="cmyk")
image = image.ifthenelse(cmyk, 0)
image.write_to_file(sys.argv[2])

I used python since it's very simple, but any language with a libvips binding would work. Are you on windows? C# would also be a good choice.

I can run it like this:

$ time ./spot.py "spot (CMYK:10,20,30,40).tif" x.tif[compression=deflate]
real	0m0.551s
user	0m0.887s
sys	0m0.152s

So it's taking about 0.5s for a 6k x 4k image and makes this:

A four-band CMYK image with 0 in most pixels, but 10, 20, 30, 40 in pixels which were 1 in the half-tone.

I think I might have misunderstood. You are splitting the PDF into a set of separations, each of which will be printed with a different PANTONE. I suppose this means you could have any number of separations, is that right? So really you need to approximate overprinting too.

I would turn each halftone image into CMYKA (CMYK with alpha), then composite however many separations you have into a final CMYK in print order to simulate overprinting.

Maybe:

#!/usr/bin/env python3

import sys
import re
import pyvips

# parse spot colours from all input filenames
layers = []
for filename in sys.argv[2:]:
    match = re.match(r'.*\(.*CMYK:(\d+),(\d+),(\d+),(\d+).*\).*', filename)
    if not match:
        print(f"CMYK specification not found in {filename}")
        sys.exit(1)
    spot = [int(match.group(i)) for i in range(1, 5)]
    
    # load the image, make a solid block of the spot colour, use the 
    # halftone to set the alpha
    image = pyvips.Image.new_from_file(filename, access="sequential")
    layers.append(image 
                  .new_from_image(spot)
                  .copy(interpretation="cmyk")
                  .bandjoin(image))

if layers == []:
    print(f"no layers specified")
    sys.exit(1)

# make a black (no ink) CMYK image to use as the background
background = layers[0].new_from_image([0, 0, 0, 0]).copy(interpretation="cmyk")

# composite all layers
image = background.composite(layers, "over", compositing_space="cmyk")[:-1]
image.write_to_file(sys.argv[1])

If I make a second separation like this:

$ vips extract_band ../theo.jpg mono2.tif 2
$ vips relational_const mono.tif spot2\ \(CMYK:40,30,20,10\).tif[bitdepth=1,compression=ccittfax4] moreeq 50

I can run:

$ time ./spot.py x.tif[compression=deflate]  'spot2 (CMYK:40,30,20,10).tif'  'spot (CMYK:10,20,30,40).tif'
real	0m1.329s
user	0m12.126s
sys	0m0.786s

So it's taking 1.3s now. The final image looks like:

You can see the separations are rendered in order as background, spot2, spot.

I want to thank you for your elaborate answer.
I was unable to study further due to time related issues.

My use case is that I have a pdf that contains CMYK and on top of that any number of PANTONE colors. I intend to obtain 1 bit separation (halftone) tiffs for every channel, Cyan, Magenta, Yellow, Black, Pantone#1, Pantone#2, Pantone#3, etc. My intent is to recombine all the channels into a single PNG color image, where I know that some colors will overlap, and the final results will have to be as close as possible to a printing press, that prints with plates and has a separate mechanical unit for every color that is printed. The plates that are used can only contain a single color.
I will have a look further into the pantone issue, but for now to keep things simple, being able to use a C, M, Y, K set of values for a certain file should not be an issue, as long as I can interpret a 1 bit halftone image with a certain C,M,Y,K value to simulate colors other than Cyan, Magenta, Yellow, Black, should suffice.

I will come back with questions regarding your answer, as I can only now study it :( Sorry for delay