Untitled Document

I have provided in the FILE section of the group, 2 screenshots (jr redderblu and jr lessred). When dealing with deep blue skies on a printing press, what is the best I can hope for? Do I try to minimize red and up black, sacrificing the rich blue in the original RGB image. I do not want a purple sky? Cyan will not be rich enough, right?

The image is copyright protected by SuperStock and I use it here only for instruction in hopes of learning for my client.

If this is OT, please contact off list.

John Robinson

___________________________________________________________________________

Hi John, purple is one of the most difficult colors to print, as well as 50% gray's and browns - but why? The reason is that any values near 40-50-60% in a tone curve increase in dot gain - the most on press. The "standard deviation" on a press is =/- 2% midtone = (50%) gain. This means that a 50% Cyan and 50% Magenta are always 'changing' in color. The secret is to print and control the midtone gain. Points further down the tone scale 20% or 80% won't change on press, it is the apex of 50% that changes the most.

If you go to <www.gain.net> under 'process control' download a pdf titled "Process Control Reference". It will show these changes.

Dan Remaley PIA/GATF

Senior Technical Consultant

Process Controls

412.259.1814

___________________________________________________________________________

John, I don't think that purple is an issue here in SWOP type conditions, in your edited example (I agree about the original). Purple often happens once the magenta goes over 70-75% with high cyan values near to solid. Safter values of magenta would be 40-60m with 90-100c ranges (depending on the artistic aimpoint) if wishing to avoid any chance of purple/mauve and with lower cyan values one may have to be more conservative in magenta. Also one may wish to reduce or remove yellow from the deeper blue tones (lighter blues should not end up with yellow values of any great concern).

Stephen Marsh.

___________________________________________________________________________

Stephen and Dan, thanks, so realistically, the richness of the deep blue in the original is impossible in CMYK? The client needs to know this going into the job. Even with all the understanding of channels, curves, etc. Not that I am anywhere near up as most of the list.

John Robinson

___________________________________________________________________________

John, I think that this is a common situation. The original blue is not really OoG for TR001 type conditions. This has been covered in past threads on the list (when blue turns purple), but simply put the default conversion offered by common profiles leaves something to be desired from a human observer - despite the CMYK LAB value being close to the original RGB LAB reading the human observer perceives the CMYK version as more purple if left with the default magenta values.

If I have to try for a 'good' blue in conditions similar to SWOP CMYK, such as when a client would have chosen Reflex Blue or Blue PMS 072, or MS Office 0r0g255b - I usually go with a value of 100c70-75m0y?K depending on various factors.

The original RGB converted to three common SWOP profiles gives the following readings for the same area of sky (rated from good to bad in terms of magenta value):

Default Custom CMYK: 99c80m14y7k

Chromix TR001: 98c86m7y18k

SWOP v2 Web Coated: 100c91m19y6k

In all cases I would either edit the RGB or edit the CMYK so that the magenta value was 75% max or 5-10 points lower for the deepest blue area, as none of these default conversions are good in my opinion and are too high. Also note the blue killing yellow values in the various conversions, which although colorimetrically correct is incorrect in the artistic interpretation of making the best possible blue in CMYK.

All blues that are not solid cyan are problematic in CMYK, the deep blues are just really lacking when compared to RGB. Lighter blues can also be an issue being too mauve/purple and the magenta to cyan ratio is even more critical than in the deepest blue when one is working with lighter blues.

It is always good for all parties concerned to have basic knowledge of the limits of the process at hand and to have realistic expectations. The whole process is one of diminishing returns, the final printed result is always far removed from the original. When viewed in context, there is no problem with a good CMYK version of a difficult colour - it is only when one has a side by side view of the scene/capture next to the CMYK print that one is really disappointed.

Having a range of sample swatches offset printed or less ideally proofed to offset standards can be helpful. When editing images where it becomes an issue of making artistic calls on what a hard or non reproducable colour should be one can look at the chart and see what values max out at blue before things look purple. Then, despite what the original image and the conversion does, you edit things to be safe to what you know works.

Regards,

Stephen Marsh.

___________________________________________________________________________

On Oct 13, 2006, at 3:19 PM, John R wrote:

Stephen and Dan's posts point you in the right direction with regard to the purple sky problem. Communicating this to your client may be the more difficult problem. Sometimes, the best thing is for the client to have never seen that beautiful, deep blue sky in the RGB image. If they have seen it, you might suggest to them that they had best forget they ever saw it.

Henry Davis

___________________________________________________________________________

Over the years I have tried a variety of techniques using both Photoshop and 3rd party plug-ins for masking hair out of a white background. I currently have about 60 studio images to mask and am looking for group input on accurate and efficient techniques some of you may have developed or tried.

I have used a variety of techniques over the years, but have found none that are both accurate and time-efficient. The final masked images will be placed on a variety of different color/density backgrounds.

I have posted a sample image at: www.digi-graphics.com/ctheory.htm.

If this is off-topic, please respond off-list to me at: com

Thanks in advance for any suggestions.

Les De Moss

DigiGraphics LLC

___________________________________________________________________________

Google Russell Brown Advanced Masking...its a free video tutorial that is the best I have seen.

___________________________________________________________________________

Make as best a mask you can from the Yellow channel (Blue?) Paint the edges [not the highlights] with a light tint of the background (darken blending mode?) Keep the backgrounds as light as acceptable

Laurentiu Todie

___________________________________________________________________________

Les, I have no time to look deeply now, but it appears that you could do yourself a major favour by experimenting with the use of the various matte commands after you have true transparency (no layer mask) - Remove white matte and or Defringe would be required.

This is above and beyond whatever "perfect" masking method you decide upon.

I have many ideas as this has been bread and butter work over the years for me, I will try to list some later (but perhaps Katrin Eismann covers all of them in her "making and extraction" book anyway!)

Best,

Stephen Marsh

___________________________________________________________________________

Les, that's an interesting problem. It's doubtful that you're going to find any single method or plug-in that will make this kind of masking efficient if you're talking about doing it in several minutes and coming out with an outstanding selection. As for accuracy, wisps of hair do not require accuracy, only naturalness and realism. Here's a solution that enabled me to make a satisfactory (to me) selection in less than 10 minutes. The time required will depend on just how good you want the image to look and on what kind of background is going to be lying behind the hair. I moved my selected image over a previously posted image of the interior of an office area that had large, dark ductwork hanging from the ceiling. The white (or pale) fringes along the edge of the selected image were not excessive, but they were pretty obvious in front of the dark areas, less so in lighter areas of the image.

To make my selection, I made a copy of the Blue channel and used Curves to enhance its contrast, but not very much. Oddly enough, when working with delicate strands of hair, a strong mask is a hindrance and not a help. It worsens the white or light fringing. Of course the interior of the hair and the girl's face and shirt contained considerable detail that you don't want in this kind of mask. So I painted out the interior with black, avoiding the edges and the wisps of hair. No effort was made to be precise, just to get rid of the interior detail. Inverted the channel copy and made a selection from it. Used this selection to create a Layer Mask for the girl. Deleted the bulk of the white background by applying the mask. Changed the image resolution to match the image into which it was to be pasted. Draggged the selected image to the target image. Created a Layer Mask for the selected image layer. Changeed foreground color to black, brush blending mode to Darken with an opacity of about 25%. With the Layer Mask active, carefully painted over the edges of the hair where the white or light fringes were showing. If you overdo it, change the foreground color to white and the blending mode to Lighten and repaint the over-edited areas. With patience, perhaps some variations in brush opacity, brush size, etc., and care you will have an image where the delicate wisps and edges of the hair blend naturally into the background. I don't recall what Russell Brown did, but I seem to recall that both he and Katrin Eismann created a border selection around the perimeter of the hair and used Multiply mode to blend the edges of the hair into the background. It was a nice technique, but the result was not all that satisfying, not to mention the amount of time it takes to make a good selection of the edges. Those who don't like Layer Masks are invited to find another way to accomplish this. There must be other ways, and I for one would surely like to know what they may be. Let us know what works for you. This kind of problem really stimulates the old thinking process, and posted solutions do the same.

Howard Smith

___________________________________________________________________________

On Saturday 07 October 2006 13:39, john castronovo wrote:

I was curious about this, too. Why *does* LAB use green/magenta and yellow/blue? Could other axes be used?

Bjorn Helgaas