Where to begin when touching up a photo?

[Tom_T...@adobeforums.com]

I have an old photo (from the U.S. Civil War, i.e., really old) that I need to touch up. It will be printed eventually. I need to know what is the best order in which to perform various operations.

It is a tiff, 105MB, about 40 inches square, 700 ppi, 16-bit black and white.

What I need to do is reduce the size to about 17 inches with 300 ppi, clean up some cracks and spots in the photo, sharpen the face of one person who is a little out of focus, play with the lighting to bring out a little more detail and finally, turn it into a duotone.

And, of course, do all this non-destructively so that I can make changes in the future.

I know how to do all these operations, but in what order is best?

I have gone ahead to clean up cracks and spots while the photo was at its huge size and resolution. I then reduced the size and resolution. This was done "destructively." (Of course I still have the original).

It seems to me that the last thing to do is do make it a duotone because, in order to do so, I have to make it an 8-bit grayscale, thereby loosing some detail. So this should be the last thing and I can make multiple duotones.

Is this right? And what is the best order to do the sharpening and highlight/contrast? I will use adjustment layers so as to not destroy any pixels.

Further, is there a way to create a dutone by using adjustment layers?

Lot of questions! Any suggestions are much appreciated.

Thanks,
Tom

[Ramón_G_Castañeda@adobeforums.com]

First of all, do you still have access to the original photo?

If so, scan it again to the target size you want for printing. You'll get much, MUCH better results than if you downsample that monster scan you've got right now.

a tiff, 105MB, about 40 inches square, 700 ppi,

What you have right now is a whopping 28,000x28,000 pixels image.

All you need is 5,100 x 5,100 pixels to print 17 inches with 300 ppi.

[Lundb...@adobeforums.com]

That would be an industrial strength scanner I guess.

[Tom_T...@adobeforums.com]

Sorry, all I have is the tiff.
Tom

[g_ba...@adobeforums.com]

try doing as much retouching as you can on new layers...save new .psd versions as you go along so all your time is not in one file...the most accurate retouching will always be ay 100% actual pixels...

[barke...@adobeforums.com]

I would return to the original as suggested and rescan. Then first thing for me would be to sort out the white point, black point then also check in the levels window to see where most of the tonal information is. I would probably be looking to steepen the curve in the tonal range that includes the facial features.
Then I would move on to non global issues, however I usually leave the sharpening until the end, but I know many who prefer to do the sharpening after they've dealt with the tonal balance.

[pfigen]

If you've got a 40 x 40 700 ppi and it's only 100 mb or so, it has to be a one bit bitmap image. A 16 bit Grayscale of those dimensions would be almost 1.5 Gigs.

The first thing you have to do is determine what you really have here. You won't be able to retouch a 1 bit bitmap image. It will have to be converted to Grayscale first. From then on it's like any other retouching job. Do your retouching on an empty layer above the background layer. Use Curves Adjustment Layer to affect your tonality.

I would NOT downsize your original. I would do ALL my retouching on the full size original and only downsize a copy when you're ready to print. By doing all of your work on the super hi-res image, any retouching flaws will be minimized when you finally shrink your image.

[John_...@adobeforums.com]

I agree, do everything on a copy ( you did ) at the biggest size you can. This may prove to be a bit of a challenge with my approach, however. You're going to get better results by ( IMHO ):

1.) You've done your scratch and dent removal, good. Now, if it's 1-bit, convert to GS and then convert to RGB. You can now look at which channel is the best to save in the final GS.

2.) With it still RGB, convert to LAB. You'll want to do your curve adjustments and sharpening on the "L" channel.

3.) You can also get creative with curve adjustments on the "A" and "B" channels that might help in getting the sepia ( or duotone ) look you're after.

All this can be done in 16-bit ( I think ). But, the size is an issue. I see no risk in sizing it down ( a copy of course ), you sound like you know what you're doing.

There are a few very good books by Dan Margulis ( 1.) Professional Photoshop, and 2. ) Photoshop LAB Color ) that could help you through some of the steps. Good luck, sounds like a nice project.

[Adria...@adobeforums.com]

As you make complex selections throughout the process, save them. And as per g ballard above, convert to a psd and work on new layers. Then you can (after turning off those between your current working layer and the unaltered pix layer) turn the top layer on and off to, in effect, view just the changes you've made. I catch a lot of mistakes that way. It's also a great way to impress a client with the final version vs. original.

If you have already worked on the background layer (but haven't cropped or straightened the image) double click on the background to make it a regular layer; open the original scan and hold the shift key down as you drag it into the image you're working on. Maneuver it to the bottom of the stack.

[pfigen]

This is a grayscale image if I'm not mistaken. No need for L*a*b in this case. That would only be of help in certain color images. They didn't have color images yet in the 1860's.

As far as file formats go, psd might be too limiting by the time you have layers. It's very likely that you'll go over the 2 gig limit and need either layered tiffs or psb's. What you really don't want is to encounter the somewhat rare situation where a psd files saves but ends up being too large to re-open.

If you're working on Adjustment Layers, any masking will be automatically saved within those adjustment layers. The original poster mentioned the level of work needed in this image and it's unlikely that it will need any really complex masking to fix cracks and sharpen a single face.

Part of the job of effective retouching is identifying what is really necessary and what is not. I'm much more in favor of simplicity and working smart whenever possible. This does not need to be more complicated than it needs to be.

[Tom_T...@adobeforums.com]

Hello all,

Thanks very much for all the suggestions. There is a lot to ponder and experiment with.

I was a bit wrong in describing the image. The original is 1.4GB (only ten times bigger than what I said!), 27625 by 28170, grayscale 16 bit tiff.

If anyone is interested you can see the photo at:

<http://en.wikipedia.org/wiki/Wikipedia>:Featured_picture_candidates/African-American_soldiers

The above has an interesting discussion of the photo.

To get the original go to the Library of Congress at:

<http://memory.loc.gov/master/pnp/cwpb/01900/01930a.tif>

The photo is in the public domain.

Tom

[John_...@adobeforums.com]

LAB wouldn't be so much for color, but isolating a mask channel for retouching. Perhaps you are correct in that I may be over complicating it a bit. I only throw it out there as a consideration, nothing set in stone. If you can get a copy of Photoshop LAB Color, check out pgs. 226 to 229. It's cool, just keep an open mind and look at all the possibilities that could make the job go quicker and easier, in some cases.

[pfigen]

Tom,

I downloaded the hi-res you linked to. That was a 16 bit untagged Grayscale tiff that was 105 mb. It appears that it was scanned at 2125 dpi, which is way overkill for recording detail on any print. It does, however, give you fine resolution on all the flaws in the image.

I think, in this case, due to the extreme amount of cracking and other artifacts visible at 100 percent on this scan, you might in fact be better off reducing it to the size you are reproducing it, or at least to the largest size you might ever reproduce it and see if that doesn't eliminate some of the work for you right off the bat.

The first thing I would do is run the Shadow/Highlight filter, concentrating on the highlight portion to bring some more usable detail into the highlights.

After that, it looks like a matter of cloning/healing until you're happy with the results. There is a lot of fine detail that *could* be cleaned up. It all depends on your patience. One could easily spend a couple of days on this image alone, maybe more.

Technically it's not hard, just tedious.

Good luck and take plenty of breaks.

Peter

[Buko]

I'd adjust the mids so they are a bit darker save the original.

now save a dupe of the file and convert to 8 bit I see no reason to work on this image in 16bit.

[Lundb...@adobeforums.com]

Your first link goes to the Wikipedia entry for, guess what?, Wikipedia.
The second link won't load. So much for curiosity.

[Ramón_G_Castañeda@adobeforums.com]

Here's Tom's first link <http://en.wikipedia.org/wiki/Wikipedia:Featured_picture_candidates/African-American_soldiers>. The forum doesn't like URLs with colons in them, so it breaks them.

Here's the thumbnail of the image:

The second link is a lost cause (memory.loc.gov). The server appears to be on dialup and it's projected to take 34 minutes to load on a DSL connection. :/

[Ramón_G_Castañeda@adobeforums.com]

The original image size at that location is 3000x1789 pixels. I've uploaded a 1024x641 pizel downsampled image with minimal correction to Pixentral:

<http://www.pixentral.com/show.php?picture=1LZCLnLUyBwi43QEsbf1GSWTyxR8m0>