:-[I am coming along on the photo; there are still numerous areas that need touching up and I would appreciate some suggestions.
(http://i90.photobucket.com/albums/k270/philbach/OPRII/LawsonE7_4_4x6.jpg)
This is where I am now. I had to paint several areas that were spotted out so to speak. Perhaps the eyes are too bright/startling.
The gray point of the original is in the background in the top left area.
:'((http://i90.photobucket.com/albums/k270/philbach/OPRII/5.jpg)
Hi Phil,
Your spots look like a perfect situation for frequency separation. Jonas, help me here with a reference link . . . If you used that technique you'd end up with a restore that looks more like a photograph and less like photo art. I've started using frequency splitting on almost all my restorations - I use it on a grayscale luminosity group and deal with the color later, once the luminosity is restored.
Lurch, good point about frequency separation. I just learned how to do that and tried it here with poor results. Lots of the spots were in the High frequency layer. When I worked on it strange things happened. What I forgot to do was to limit the clone tool to the single layer I was working on. Ha. Thanks
Phil,
The one thing I notice is that it looks like you have replaced the boy's eyes? In the original it looks like the he is looking to your right, but in yours he is staring straight on.
It will be interesting to see how this works out.
Margie
Quote from: lurch on October 19, 2010, 04:40:17 PMJonas, help me here with a reference link . . .
Here you go :-): http://www.operationphotorescue.org/forum/index.php/topic,2687.0.html or for the original thread check this: http://www.modelmayhem.com/po.php?thread_id=439098&page=1#post9585669
I'd suggest you use a pretty low radius for the separation (just so low that the dots start to disappear), paint in the color on a new layer in between on "Color" mode and use the clone stamp or healing brush on the high frequency layer to get rid of the spots.
Probably a little smudge could be used as well where nothing else works, but in most cases it's imo best, to repair and not to rebuild most of the images :-).
PS: Using a high pass with huge amount from the original image combined with a medium amount Gaussian blur on the resulting HP layer might help to get some of the original's shape back after repainting if applied in "Soft Light" mode to the restore.
Well the separation technique hasn't worked well for me yet. Interestingly the area that responds to correction is the High frequency layer more so than the GB low frequency layer.
Margie, I agree about what you said. I didn't replace the eyes but I did draw in the iris and pupil out of necessity. The end result is as you described. A very small error on my part makes a big difference on the eyes.
I guess I'll be back.
Der it takes a fairly aggressive GB to cover up those big blemishes.
Thanks everyone for your suggestions.
Phil,
Looks to me that your color is washed out in your restoration. The highlights on his face don't seem to have much definition. The shadows are disappearing. His red shirt looks like it is bleached out. I downloaded your original image and did a levels adjustment moving the black and white sliders in the RGB layers. To me the color seems better, and doesn't look as washed out.
(http://img820.imageshack.us/img820/2953/lawsone744x6copy.jpg)
I see what Margie is saying. When you did the pupils there it too much white of the eyes showing now compared to the original. Your pupils need to be larger along with repositioning.
Johnboy
I like Johnboy's levels adjustments. As for the eyes, I would zoom in 200% or more or work on pixel level to repair them. The blue channel may be of some help to see what your repair should look like. I often place the 2 photos (original and WIP) parallel next to each other and look back and forth frequently while I do the repairs.
It would be great if most of the work in important features like eyes and mouth could be restored rather than repainted, no?
Hannie
Hannie I agree. I am going to rework the photo. The eyes turned out wrong but it didn't become obvious until I finished what I was doing and zoomed out. But I will head back to the drawing board so to speak and start over.
I tried various flavors of frequency separation. In all cases the Majority of the damage was in the High Frequency (Texture Layer). Patching that layer blurred the photo a lot. However I did do it to some parts of the photo.
I worked on the eyes which did seem to be aiming too much straight ahead so I feel that progress has been made.
(http://i90.photobucket.com/albums/k270/philbach/OPRII/Wed2.jpg)
I like the color in the latest version much better. To me the highlight and shadow along the jaw line (viewer left) looks quite sharp. Also along the chin line. Can that be soften some?
Johnboy
Good Point, Johnboy I will do that.
Phil,
I think you are doing a great job of repairing the damage and getting rid of the spots, but if I compare your latest version with JohnBoy's (color corrected version) in the blue channel; I see differences. The boy also has teeth showing at the bottom.
(http://i192.photobucket.com/albums/z10/hayesbucket/Lawson_blue.jpg)
I feel like a real party pooper, but I have a problem with the color correction. I feel like JohnBoy's is closer to the original color than your latest. I think the colors look muted or dull. I'm not sure if this happened by doing the frequency separation? Or is this a case that you went with your original color correction and if so, by what method did you use? If I go to the blue channel, the white point looks like it should come in. However, because it has already been corrected you can't move the slider now without making it look bad.
I was curious as to whether the boy's hair is really as red as JohnBoy's Level's correction, so I called the owner of the photo and her husband said she would call me tomorrow. I also did a Curves Adjustment and came up with the same results.
Margie
Hi Phil,
I fiddled around with your original copy a bit to demonstrate for you how effective frequency separation can be, once you get the hang of it. Here's my CS5 workflow, more or less.
1) Do a color correction. In this case I used levels on a channel by channel basis.
2) Add a black and white adjustment layer using default values
3) Merge all visible layers to a new one and make a copy of that new layer. Name the two blur and hi pass (or texture or some such). Turn off the top (texture) layer, the black and white adjustment layer, the background layer, and the color correction layer.
4) Use Gaussian blur on the layer labeled blur with a radius that just gets rid of what you want to remove. In this case I used 4 pixels.
5) Use Apply Image to apply the blurred layer to the layer named texture. Set the mode to subtract, scale to 2, offset to 128. Leave other parameters alone. Set the layer mode to Linear Light.
6) Group the blur and texture layers, label the group, and set the group's blending mode to luminosity.
7) Do your cleanup on the texture layer, with whatever tools you prefer (clone stamp, healing brush, context aware healing . . .) using a good area as a source. For that process, set sampling to Current Layer. I like to clip a normal mode cleanup layer to the texture layer, but to do that you have to turn off all the other layers, set sampling to All Layers, and sort of work in the blind, on the squirrely gray Linear Light layer. You might have to clean up the blur layer also, if there are undesirable variations in luminosity as show up here.
8) (This is supposed to be the numeral eight followed by a left parenthesis, but it keeps being turned into a smiley.) Turn on the background and levels layers. Color the stains that remain on a color-mode layer above the luminosity group. I use a 100% brush set to color mode and sample a nearby good color.
This is the basic recipe; as you get experience you will modify it to taste, and to add in things like dodging and burning, etc. For my fiddling, I didn't do anything on a color layer - just went far enough to demonstrate cleaning up luminosity.
(http://i206.photobucket.com/albums/bb314/lurch_album/ScreenSnapz-2.jpg)
And Margie, FYI, frequency separation doesn't affect color balance in any way.
Hi Phil,
Your damage repair looks great! I always have trouble with hair, can never make it look right. ???
This topic is so interesting and so much can be learned from it. Lurch, thanks for the step by step tutorial. I know Jonas made a tutorial on this method but it never hurts to have it explained a little more.
I agree with Margie that the colors/toning are still not right. Below is the part of the OPR Handbook that deals with color correction: (can't hurt to have a look at it again ;-))
(http://i210.photobucket.com/albums/bb18/marijtje2/OPR/levels1-1.jpg)
(http://i210.photobucket.com/albums/bb18/marijtje2/OPR/levels2-1.jpg)
This method works on most images, it is easy and gives a natural result. It work particularly well on the this photo, from left to right: original, levels as by OPR Handbook, Phil's version:
(http://i210.photobucket.com/albums/bb18/marijtje2/OPR/Phil.jpg)
Hannie
Lurch, thank you so much for taking the time to go through a step by step. I was familiar with the frequency separation but what you showed me was placing a black and white conversion adjustment layer which has allowed me to proceed. Thats what I am working on now. I used a higher GB then you did probably because the file size of the original photo is larger than what is posted here. This will take a while. But so far the results are fine.
As for color adjustment, I found the gray point to be in the background all over the upper left corner of the photo. So is the background light blue or various shades of gray? When I used channel by channel levels, the resulting photo was oversaturated and had a background that is light blue r=191,g=206, and b=217 So I went to lab and corrected the gray point to a=0 and b=0. That resulted in the gray. I agree that that the color isn't quite right and I'll post and update later
(http://i90.photobucket.com/albums/k270/philbach/OPRII/Screenshot2010-10-21at73547AM.jpg)
Thanks to all for helping me. Hey this is an interesting topic for me.
Phil, LAB can help in lots of situations, but what you are trying to do is to force the photo to a color that it wasn't originally. Plus, if you get good results go with the flow, but if you make the photo's colors off, then don't use that method. I looked through my Photoshop LAB Color by Dan Margulis to see if he did as you were doing and could find nothing, although I did see the idea posted online. I think where you are going wrong is that once you find what you want for the closest thing to a neutral gray; you then go into LAB and set your a=0 and b=0 which then creates havoc with the photo's color. Yes, you will find shades of gray all over the photo as that will influence the shades of colors--more gray, darker shade of color. By going in and changing the a and b to 0 you made the background totally gray which it is not.
QuoteIn Dan Margulis book Professional Photoshop The Classic Guide to Color Correction (5th Edition) on page 213 is some good advice:
"Don't get into the habit of forcing things to be a specific color unless you're certain that you know what that color is."
The owner of the photo said she would get back with me, but she did mention that the picture is a school photo. That would make the background a professional backdrop and I would say that it is blue with variation throughout out. If I can find Mary's tutorial on repairing studio background, I will post the link.
Margie
Well I am going through Dan's Book for the second time. Ha, but I am not too worried yet about the end color. I almost always start with levels on the channels like you suggested but the end result wasn't so good when I did that initially. So I sort of drifted off. I went with the assumption that the gray point was gray and therefore the whole background would be gray. Not so I guess.
However I feel comfortable about getting the color pretty close to correct eventually. I'll post some updates in a few days.
I am now working with Lurches tutorial on frequency separation which helped me work with that technique.
Phil, if you pick a gray point from an area of a damaged, uncorrected photo and you are not sure of if it that area was gray in the original undamaged photo, you are going in the wrong direction.
It is for that reason that we always ask to do a basic levels in RGB channels first, it is the only safe way to work from and you get no weird results.
Once you use the grey dropper in the wrong spot and work from there it is impossible to fix the colors in a satisfactory way at a later point in your restore.
Hope this helps,
Hannie
Phil,
Per the owner, the boy's hair is brown with a hints of red throughout. The background is blue in color, but with variation throughout--do not make it a straight blue. I know this photo is going to take some time, but one thing to be very careful of is the blurring I see. What you do not want is to have the child's face to look like a painted mask. It never pays to blur the eyes, lip, teeth, or anything that will catch the the viewer's eye as looking off. Had this photo not had those details, it would be a different story. Check all the channels for missing details.
Besides what Hannie has added, getting the gray balance close is a trick photographer's use all the time. That is why you will see a gray card in a test photo. We don't have the card, but are trying to find a close neutral gray to use for the correction only. That is where a Curves or Levels Adjustment comes in.
This is a very interesting photo in that there are different ways to approach restoring it. That is why I like Katrin Eismann's book Photoshop Restoration & Retouching.
Margie
Thanks for the info, Margie. I will post a redo tomorrow. The color I have is as you described. I am have read Katrin's three books but right now I am in Wisconsin and the books are in SC. I have used frequency separation as described by Lurch and others. I have checked the channels and have placed the lower teeth in the redo. So its quite different from the last post.
But I appreciate the information you posted. My background is bluish and I did use the channel by channel levels at a mild setting which seemed to have worked ok.
Well I am getting to the end. I used the blue channel fairly extensively to bring out his lower teeth. Levels on Channels for color.
(http://i90.photobucket.com/albums/k270/philbach/OPRII/Boy.jpg)
Phil,
Here is what I have got so far and am only about 75% done. There is still lots of damage (plus discoloration under chin) showing, but see what you think.
(http://i192.photobucket.com/albums/z10/hayesbucket/LawsonE_WIP7.jpg)
Margie
Wow Margie, you really have come a long way with this photo. From my perspective your version looks fine as far as the spots are concerned. Far more contrast and sharpness than in my version. Perhaps their is a slight yellow tone in the facial coloring especially around the nose, but that really looks good.
So how did you remove the spots?? Ie frequency separation or other methods?
I think his hair on his left side (our right) doesn't go straight down but sort of parts near the middle and the left side drifts to the left. I like your eyes more than my eyes and perhaps one reason for that is the pupil on your version is less obvious than on mine and more realistic appearing too I might add.
I'll post another version Monday.
Phil,
Thanks. No, I did not use the "frequency separation" method to restore. I found neutral gray by using Dave Cross' method and then used the gray eyedropper in Curves (Levels would have worked also). After hearing the boy's hair was more of a brown/red; I found the ideal gray point was on the buckle of his suspenders. After making the color correction, I made a duplicate copy and then took the blue channel (best detail)>Ctrl (Cmd-Mac) A (all)>Ctrl C (copy)>and then Ctrl V (paste) on the 1st photo that is in color. I then went to Mode and changed to Luminosity. This is a trick that Kiska has illustrated several times on the forum. What I am noticing in your way is that you have blurred the entire photo--Everything which you don't want to do!
When I change the mode to Luminosity, I noticed that the boy's red sweater didn't look as good as before, so I added a layer mask and painted that part back in. My next move is to do a Merge Visible and duplicate that layer. I do that so I have a copy I can go back to. From there it is a matter of using the healing brush, clone, lasso, patch tool, and many layers and layer masks. I did a Gaussian Blur, but only for the skin and a touch on the hair. When I do a blur, I do just enough to get the effect I want and then I do a layer mask and conceal all. Then I go up and pick a soft brush and reduce my opacity to about 65% (varies) and then on the layer mask paint back in the skin. I worked the eyes by working with colors there and using them to conceal the dots. I try to stay true to how the eyes are and the dots are brushed in with like colors. As mentioned before, I never blur the eyes as that would give a dull look.
Margie
Great topic!
Margie, your version shows that less can be better than more. The eyes and mouth look just like in the original, there is a little damage left but that is a small price to pay for having the image stay true to the original.
Thanks for reminding me of Kiska's method, I should be using that more than I have been.
So much to be learned here!
Hannie
I agree, Hannie and thanks Margie. I will go back and work from the start again and when I am getting close to being finished. I will repost. It will be a while since I need to be off line for about a week or so.
Phil, do you have CS5? One thing I discovered with my new Photoshop is that with a varied background like the one you have here, you can make great use of the Content Aware healing brush. I went over any area of the background that looked "iffy" with the brush set at a fairly large size, and the colors and "off" areas blended into a lovely mottled blue within minutes.
I'm finding that while Content Aware isn't the answer to everything, it helps in places one might not expect. I was able to eliminate spots on the boy's face that were surrounded by good skin by making my brush just a little bit larger than the discolored area. It kept the skin shade that was there and the spot was gone. In the case of your little boy, it's not applicable everywhere, as there are too many yellow spots virtually on top of each other. But if you can open up some good skin areas, then you can use the patch and healing tools, as well as the clone tool to get good results.
Content aware is tricky, because as with patch and heal, you have to stay away from or completely surround areas of high contrast. But I've been very happy with the effects I've gotten on the several restorations I've completed and the one on progress right now with judicious use of content aware. It does a beautiful job on most texture or at least gives you a good start.
Also, one thing I wanted to mention is that I'll sometimes choose a black or white area to use as my neutral point. Many times that works as well as looking for a gray if you aren't going to use the full Dave Cross method, which is a little more complicated.
Margie, I'm curious as to what methods you used to get the wonderful result you did.
Have a great week, all.
GK
Thanks GK, look at my last post on this thread as I go through the method I used.
Margie
Sorry for the long time for a response, but I've been offline for more than a week. I tend to migrate south in the winter.
Glennab, I do have CS5 and have messed around with content aware fill and healing and do like it.
However in this latest rendition. I started over using Margie's recipe which she discussed earlier.
Levels on all three channels
After that copy the blue channel and paste that into the layer above using luminosity as the blending mode and a layer mask to mask out the bad stuff.
I then use frequency separation as described by Lurch to work on the spots
Finally I fiddled with the color some.
(http://i90.photobucket.com/albums/k270/philbach/OPRII/LawsonE7_4_4x6-1.jpg)
Phil,
Big improvement!
Margie