Legacy ImageMagick Discussions Archive

I am trying to programmatically convert CMYK images to RGB.
If an image contains an embedded profile then I can convert to RGB like this:
convert source.eps -profile /path/to/sRGB.icm dest.jpg, which is great.

The problem is, if it doesn't then, according the the IM examples, the first use of -profile will be used to mean the profile of the source image, i.e. I should specify a second -profile flag to specify which profile to convert to. I guess this is the reason why the results look so bad.
In this case, using:
convert source.eps -colorspace RGB dest.jpg seems to work much better.

My questions are:
- What does -colorspace RGB do exactly, i.e. which color profiles are used with this option? Why/how does this seem to work reasonably well for images that do not contain an embedded color profile?
- What's the best practice for converting a CMYK image to RGB if the source color profile is not known?
- What's the easiest way to determine if an image does in fact contain a color profile, i.e. so that I then know which of the above approaches to take?

Any help much appreciated!

Thanks,
Martin

I had this same issue a few months back. I was looking for a way to determine the actual
name of the color profile so my script could choose the appropriate profile for conversion.

I got no response to my post though, and never found the answer myself.

In the end, I found that if I used the more generic profiles, one for CMYK and the second
RGB, the results were fine. I suspect that if the color profile is not important enough to
be known, it's not important enough to be specified exactly.

Thanks Andrew. I came to the same conclusion (regarding generic profiles) and ended up doing it like this:

1) Determine if the image has an embedded image profile by using:
convert image.eps profile.icc
If this command doesn't return an error code, and the file profile.icc is created successfully, then the image contains an embedded profile.
[Surely there's a better way to detemine this?!]

2) If the image does contains an embedded profile then use it, i.e.:
convert input.eps -profile /path/to/sRGB.icm output.jpg

3) Otherwise, use any CMYK profile as input:
convert input.eps -profile /path/to/USWebCoatedSWOP.icc -profile /path/to/sRGB.icm output.jpg
This gives better results that using the -colorspace RGB option.

This seems pretty long-winded, and inefficient (especially the first step) so if anyone has any better ideas then please let me know!

Thanks,
Martin