From: "Glenn Wilton" <[EMAIL PROTECTED]> > Thanks Mike > > I think I need to clarify what im trying to acheive. > Im actually printing with a 6 colour inkjet printer using disperse dyes > with profiles i have made. The Gamut of disperse dyes is about 70% of > standard CMYK except where we use spot colours. Therefore the RIP is setup > using Perceptual Rendering intent. This is 'safe' but not good for spot > matching.
OK. > Since I need to match Spot colours accurately you would think I would use > Relative Colourmetric intent. For Spot color matching only, I would consider abscol, which preserves the white of the source color space. > However I often have images and vectors that must use the same intent or > they would look different. (ie Flattened PDF's) So I cant set vector to > Relative Colorumetric and Raster to Perceptual. > The way we solve this at the moment is to choose an RGB colour from a test > print which is usually has a much stronger chroma than the resulting spot, > However with Perceptual compression it will print accurately. This is a > manual process of looking up colours on charts. Another possibility would be to print the same test print on fabric and match the spot color directly from that. This would save a lot of steps and give a visually accurate result, which profiles are not going to do.. > My first email was a little inaccurate. What I would like to do is convert > a CMYK/Lab colour via Relative Colourmetric intent to the printer profile, > The resulting 6 colour formula is my 'ideal match'. Gotcha. > However since the RIP is using Perceptuial Intent I need to do a reverse > lookup to find an RGB colour which when convered RGB>Perceptual>Printer > will give me the same 6 colour formula. I use RGB has a bigger gamut and > gives room to compresate for perceptual compression. One consideration is > that both Perceptual and Relative Colorumetric may not produce the same > 'formulas' So, in effect you want to "overdrive" the preceptual intent conversion, so that it ends up with the same color you would have gotten, had been able to set the RIP to relcol. > At the moment I calcualte a close RGB colour then iterate through a few > thousand combinations to find the cloest match. Simulating a B2A lut, using something like a linear search of the A2B LUT or some such in VB. > What I would like to do is perform a reverse A2B lookup but NOT a B2A > lookup as the printers B2A table only uses 5 data points vs the 33 > datapoints for the A2B Table. Did you consider regenerating the B2A by inverting the A2B LUT? I believe there is an argyll command line utility by Graeme Gill that does exactly this. Lcms will then allow you to concatenate transforms ad infinitum, so you could create what amounts to a device profile that converts from your original RGB color, to your "disperse relcol" , identity transform back to Lab, and from there to "disprese preceptual". Once the transform is set up, you use it like any other transform. Another thought would be to modify the profile used by the RIP. Replace the perceptial LUT with the relcol LUT. > Tricky? Yes, basically you have to turn carwheels to get around the perceptual intent of the RIP. Your task would be infinitely simpler and more accurate if there were no conversion in the RIP at all, and you could provide a 6 channel image, one per jet. Which leads me to a couple of words about how I would do this. The profile method described above, while mathmatically sound, won't really work well. This is because mathmatical compromises were made to match the patches that were used to generate the printer profile. Those patches, in turn, were different from the spot colors you are interested in. The end result will be a poor match to many of your spot colors. To better match the spot colors, if the number of them is not onerous, I wouldn't use profiles for preparing the input color values. Instead I would individually target each spot color as follows. First, I would print your RGB test image via the RIP, and use the swatch book to manually match each color using a light source that closely matches the final viewing conditions. Then reate a test image that contained a swatch for each spot color, print it, and verify that each swatch matches it counterpart in the swatch book reasonably well. Then I would tweak the RGB values of the swatch colors that were still significantly off until they were acceptably close. Lather, rinse, repeat. After you have a good visual match for each of the swatches, use a spectrophotometer to record these printed values, and convert them to CIELAB D50. The D50 illuminant is not critical but it is the widest standard. These measurements are not used to generate the input swatch values, but they can be used for process control. If the images are photographs that include spot colors, then I would add additional "pseudo spot" colors for neutral gray, a skin tone, foliage, and other important colors in your images. For each image to be printed, check each spot color manually, including the hue (not brightness) of each pseudo spot color, and use curves to set it to the value from the swatch image. If the number of images is large, I would use the same curve on all of them, or have several curves, one for each category of image. Mike Russell - www.curvemeister.com ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user
