Hi everyone,
As promised, here is our new group challenge:
(http://img217.imageshack.us/img217/4080/millerl20164x6.jpg) (http://imageshack.us/photo/my-images/217/millerl20164x6.jpg/)
Uploaded with ImageShack.us (http://imageshack.us)Uploaded with ImageShack.us (http://imageshack.us)Isn't she a beauty?
One of our volunteers suggested that we break up the restore in parts and start with color correction, good idea!
You can post your work here in this topic. It would be great if you can also add a small description of what you did to get your result. Of course you can add comments to other posts, ask questions, joke around a bit, etc.
:loveit:
Hannie
PS
I noticed that Photobucket changed the size of my upload, originally this is a 4x6 in, 300dpi photo but now it is smaller. Anyone have a suggestion how I can get the proper size out to all that want to work it?
Update: Thanks all for the help, I changed the link to Imageshack and the image now is full res!
Hannie, PhotoBucket has the size to 14.222 x 9.472 @72 dpi, but once download it your can crop to 6 x 4 @ 300. However, I think keep it at that since volunteers will be uploading back. The only other alternative would be for volunteers to contact you or Bambi and have the file sent by either email or put in their gallery in Photo Shelter.
This is going to be fun!
Margie
Hannie
I've found that Pixentral com http://pixentral.com/ usually keeps the same number of pixels of an image. However, it can be very slow to open & process the image.
Well, I though I'd try color correcting first so I did a Levels adjustment layer followed by a Curves adjustment to make it brighter. I'm wondering if I've gone too far. What do you think?
(http://farm9.staticflickr.com/8212/8367971416_9c22e77563.jpg)
Oops. Here's on big enough to see:
(http://farm9.staticflickr.com/8212/8367971416_9c22e77563_b.jpg)
That looks about right to me.
Thanks BJTX, the download is a little too slow. I changed the download link to my Imageshack account. I did not realize they also have an option to keep size and res intact, download seems a little faster.
Hannie
Hannie, I agree about the Pixentral slowness.
WIP 1 & plans:
1. eliminated some of the aberrant colors & used levels.
2. started blocking in better color & replaced what seems to be a ghostly shape (person?) in the lower right in the original.
EDIT: WIP 1 image removed. An advanced WIP image will be posted later, time permitting.
Thanks for the interesting challenge, Hannie.
Okay, how do I download the full resolution image? It's in Bambi's gallery (!) but Photobucket won't let me download from there, even after signing in. Help!!!
Lurch, when you right click on the image in my first post you can save it to your hard drive, it is full res!
Hannie
Hello everyone,
Just joined and was directed to this challenge. At first glance I thought the image in the lower rh was a person as well, but it's actually handwriting that appears to have come from the back of another photo laying on top of this one. I made a selection of the area, flipped it, duplicated it, and played with saturation and curves to bring out some of the text. Unfortunately, most of it is just a large ink stain. I'm not sure how to attach an image here. Good luck!
Hi Betty. I think the cubicle on the far right has a divider that goes all the way to the floor. That's assuming it's like regular cubicles I've worked in. You can't tell from the damage.
Tori
@ wzthxj -- I think you are right - just a large ink stain.
@ tori -- You're right about the cubicle. My 2nd version includes a full-height division.
I didn't have much time to work on this today, but I decided to see if any more detail was hiding in the blue channel. I found some information for the two faces on the right, and for the lower hand on the right. So I copied the blue channel, clicked on the "RGB" layer in the Channels panel (window), then went back to the Layers panel. I pasted the blue channel copy above the other layers and added a vector mask by clicking on the "Add Vector Mask" icon at the bottom of the screen. Then I blacked out the vector mask by pressing CTRL + Backspace. Next I chose the Brush tool, changed the foreground color to white, and painted on the mask to reveal the parts I wanted to used (the hand and parts of 2 faces). Then I changed the blend mode to Luminosity. From here I'll start working on damage, unless anyone has another suggestion.
(http://farm9.staticflickr.com/8047/8368039233_1887556d41_b.jpg)
This really looking great Tori and thanks for the screen prints and the instructions!
Margie
I did this with the first posting so the res is low :(
My color correction matches Tori and was done the same way, curves and levels.
The rest was mostly cloning. There is still some work to be done I think. Any ideas on how to get rid of the painted look and make the face more "real" looking?
(http://i803.photobucket.com/albums/yy313/FTreeForum/December%202012/MillerL20_16_4x6_zpsacbc2d0fattempt_zps6eaa795f.jpg)
Niksmum
I just joined a few days ago and want to check on how to upload my challenge image to the forum. There are a few places that are slightly more than challenging. Needless to say, I have questions.
Judy
Judy, a good place to start is to go to The Volunteers Welcome Center as it has answers to most questions. Here is one from there: http://www.operationphotorescue.org/forum/index.php/topic,2002.0.html
Margie
Great image Hannie
I used a different approach.
Used a threshold layer to identify white and black points, didn’t worry about trying to set a 50% grey point as I have never found it to be much use.
Used levels tool to correct image with the black eyedropper set to 10 and the white to 245 so I have room to move later if I need to. Adjusting channel gamma sometime helps, not this time.
I put colour samplers on the sleeve of the front woman, on the face of the last woman, one on the noticeboard behind her and another on the document hanging from that noticeboard. Then used hue/saturation to correct colour, initially selecting a yellow from the last woman’s arm and a second from the wall below the notice board behind her. I didn’t get too concerned about the colour of the last two, the lighting for the photo is a combination of flash and room lighting so they don’t need to be completely neutral.
Did an overall boost of saturation to bring the faces back to near original
The attached image shows the changes made to those 4 points as a result of the hue/adjustment layer. I also checked the grey on the dividers and they are near enough to neutral. Only wish some of the other images we see had such a useful feature.
Suspect that ultimately a crop will be the way to go, the large area of the divider to the right will be impossible(?) to recreate realistically
(http://i151.photobucket.com/albums/s145/atholg/wip_zps7c0566f9.jpg)
Athol
Wow, this restore is coming along really well, great job everyone! :up2:
We may even be able to do this one without a crop?
Athol, I always forget how to change the black and white dropper values, had to do a search and here it is!
http://www.operationphotorescue.org/forum/index.php/topic,3071.msg27848.html#msg27848 (http://www.operationphotorescue.org/forum/index.php/topic,3071.msg27848.html#msg27848)
@Niksmum: what if you use Tori's blue channel tip to bring back texture in the lady's face?
If anyone has any questions on any of the techniques shown here, don't be shy to ask! :)
Hannie
I was wondering if this was the proper place to post a WIP. Turns out it is. Now all I have to do is figure out how to delete the copy in the other topic.
(http://i206.photobucket.com/albums/bb314/lurch_album/millerl20164x6-Edit_zpsc6f1785a.jpg)
So far I've color balanced with curves to neutralize the neutrals, set white and black points where they look good and adjusted the left-hand woman's skin (all with curves), desaturated a skoche, done a shadow/Highlights move to get a little detail in the right-hand woman's black vest, and done a local curves move to improve the right-hand woman's skin. All this was after a quick white balance in Lightroom, with a little manipulation of highlights and shadows, white and black points, and moderate noise reduction.
Thanks Lurch.
Margie
Question: Is the middle lady's hair blond or white? Do you think she's wearing glasses?
I think the hair has a touch of blond in it, and it looks like she has glasses.. I have processed this image in every way i could think of and my results are consistant with everyone Else's.
I think the woman in the middle definitely has glasses. Her hair looks like it might be blond turned gray, but there's not enough info in the damaged photo to tell, really.
I agree about glasses. Convert to CYMK and have a look at the yellow chan, she could also have some sort of a head scarf showing behind.
At least, that is what a male thinks :cool:
Athol
Yes glasses, I'm sure. Tori, the face of the lady on the right looks really good!
The lady in the middle is going to be a real battle. There is some detail in the Blue Channel but it is just bits and pieces.
(to see the Blue Channel click on the Channels tab in the Layer palette on the right and then click on Blue)
I copied the Blue Channel and pasted in in a new image of its own (File>New, change Grayscale to RGB; Edit>Paste)
On this new image I did some healing and heavy duty smoothing. I only concentrated on the centre lady, the rest is going to be hidden in the Layer mask later on.
Talking about Layer masks, if you have questions on how to use them, please ask! (I found them to be vital in the type of restores that we do)
Long story short, this is what I got, it can be pasted as a new layer on top of the restored layers of your WIP (Work in Progress). Click on the Layer and change the Layer mode to Luminosity. Use a hide all Layer mask, click on that mask and using the white paint brush bring back the centre lady's face. (black hides and white shows what is beneath the mask)
Hannie
(http://i210.photobucket.com/albums/bb18/marijtje2/OPR/challenge_zps5633df56.jpg)
Quote from: Tori803 on January 11, 2013, 12:37:03 PM
I decided to start working on the damaged faces since they scare me the most. If I can't get the faces right there's no point in spending time on the background
Tori-- Pretty much the way I feel at this point. My WIP 2 includes a some improvement to the background.
However, I've come to a stand still until I work out something for a realistic facial restoration, which doesn't seem achievable for me. I might just quit.
I enjoyed working on the background, but feel defeated with the middle face.
Quote from: G3User on January 12, 2013, 12:59:11 AM
I agree about glasses. Convert to CYMK and have a look at the yellow chan, she could also have some sort of a head scarf showing behind. At least, that is what a male thinks :cool:
Athol
Athol - I also think she has apparel on her head. Could it be something like a Nun might wear? My WIP includes it. But, that doesn't help the facial features restoration. :(
btw - your color correction :up:
Good luck to Athol, Tori, Lurch & Nicksmum -- keep up your good work !
Hannie, thanks for your help.
Had another tweak :)
(http://i803.photobucket.com/albums/yy313/FTreeForum/MillerL20_16_4x6_zpsacbc2d0fattempt2.jpg)
Niksmum
Betty, I've been proceeding slowly with the middle face. After my last steps (that I posted about earlier) I've been using Hannie's smooth copy as a guide to where her features are located. Then I started working on the neck using the healing brush and the patch tool. I also added a layer above my working layer to paint on color for parts that I'm having trouble with (trying to keep painting to a minimum.) I'm still smoothing out the line of her chin, which is a little too heavy handed. But I'm hopeful that this method will work throughout the face. Maybe this will work for you too.
Tori
(http://img194.imageshack.us/img194/5499/middleface.jpg)
Uploaded with ImageShack.us (http://imageshack.us)
Awesome Tori :up2: :up2: :up2: :up2:
Athol
(http://farm9.staticflickr.com/8360/8375825452_6bc6f27b7e.jpg) (http://www.flickr.com/photos/pixagrafix/8375825452/)
non-restore-challenge-jw (http://www.flickr.com/photos/pixagrafix/8375825452/) by Pixagrafix (http://www.flickr.com/people/pixagrafix/), on Flickr
This is an amazing challenge. I tried to keep track of things I did but I ended up with two working images and dozens and dozens of layers. I tried all sorts of techniques. Some worked, some didn't. I did a lot of cloning and healing, a little bit at a time.
This seems like one of those images that needs a reference photo.
At least I learned how to upload a photo to the forum. :up2:
Judy
Looks good Judy but Yahoo wants me to create an account before I can see the full sized repair.
Maybe another site to store the image would be more friendly ;)
Athol
Yahoo is doing something weird. I used my Flickr account a couple of days ago, but now I get a message that I have to reactivate it because I haven't used it in so long.
I can't say I like Yahoo, because it won't let you click to enlarge, but you can right click and download the image. Only problem is that the download version comes in at 6.944 x 3.625 @ 72 dpi. The other one posted was 14.22 x 9.472 @ 72 dpi. With the larger you can use the crop tool at 300 dpi and set for 5 x 4. However this one is too small and you can't go from 72 dpi to 300.
Judy, I agree it is looking good, but too bad we are having problems accessing it.
The Duck
I have always used Photobucket. Towards the end of last year they announced that they would be holding full size images loaded from that date.
The original could be recovered by downloading from the site, there is a mention of selecting size but I can't see where.
The image I loaded for the challenge was 1209 x 918 440K, what I can download is 1024 x 777 120K so it certainly been compressed. It isn't clear what size image will be displayed when you left click on a thumbnail in a post but it look to be similar in quality to the more compressed image.
Is there some restriction on OPR storage which makes this necessary Margie?
Thought you might like a new avatar ::)
You know who.
(http://i151.photobucket.com/albums/s145/atholg/ducksfeet_zps05f10386.jpg)
Athol: or He who messes with fire.
Tess will be better able to answer that question as I don't know.
The Duck
P.S. I clean up better than that! :cool:
(http://farm9.staticflickr.com/8231/8380535509_db13bda253_o.jpg) (http://www.flickr.com/photos/pixagrafix/8380535509/)
non-restore-challenge-jw (http://www.flickr.com/photos/pixagrafix/8380535509/) by Pixagrafix (http://www.flickr.com/people/pixagrafix/), on Flickr
I've reloaded the full-size image on flickr and changed the access. I don't use flickr very often so I probably made some mistakes. If this doesn't work, I'll get a photobucket account and try that.
Judy
Thanks Judy, that worked.
Terrific. :up:
Judy
Much better Judy, I was able to download a full sized copy.
Like what you have done with the RH divider and the whiteboard
Just a thought. Your restorations of the background are perhaps still a little artificial. For instance, the original wall is brighter in the centre from the flash and floor in the LH corner should also get darker to the edge. Same for the bottom of the divider
I know of two ways to tackle the wall.
Copy and heal using samples with differing brightness to best match what you can see of the original. This preserves the texture which is most important but it can be difficult to create the smooth transitions across the wide area. I used this approach for the example I posted and have a method to help with the smoothing, let me know if you are interested.
The other is to create a circular gradient on a new layer using samples of the brightest and darkest parts of the original. This will produce very smooth transitions, is much quicker but must have texture/noise added to make it realistic and that can be an art in itself.
Let me know if you would like to see examples.
Does anyone have any other methods?
Athol
I'm definitely interested in learning your smoothing method and see examples. It's always easier for me to absorb a new technique if I can spend time on each step. The circular gradient sounds like an interesting idea and one I never would have thought of. I'll have to give it a try.
Thanks!
Judy
Judy, I use two methods , either do the bulk of the repair using the clone and healing brushes or use the gradient tool to produce a new layer and add noise to match the original
In both cases I use a zig zag curve plus a levels layer to emphasise unwanted variations in levels in the result. It is easy to create, open curves and pull the curve up and down as many times as you can. You should get something like this, name and save it.
(http://i151.photobucket.com/albums/s145/atholg/ZigZag.jpg)
When added as the top layer in a repair it drastically increases the contrast of those pixels which under the sharp rise and fall of the curve. Those pixels occurring near the top and bottom of the curve are not emphasised so adding a levels layer below the curve and adjusting the centre control to say 1.2 will effectively move the midrange of the image sideways so the previously un emphasised pixels can now be enhanced.
It takes some time to get used to the effect but it is worth it. Provided you set your tools to use only on the current layer, the two layers have no effect on the edit and they can be toggled on and off independently to allow the different views of the repair. If you can't see a problem with any combination of these two layers, there isn't one.
I will use them in the following examples and apply both routinely to all my restorations, as well as helping with repairs they assist in locating damage which otherwise might be missed, assessing noise in an image or to confirm that any blending done is smooth.
The original
(http://i151.photobucket.com/albums/s145/atholg/myrepaircrop_zps1e0c8a62.jpg)
The following repair was done using clone and healing brushes.
(http://i151.photobucket.com/albums/s145/atholg/myrepaircrophealed_zps16e13691.jpg)
Zig Zag view
(http://i151.photobucket.com/albums/s145/atholg/myrepaircrophealedzigzag_zps124b31c0.jpg)
Using it has allowed me to accurately select undamaged areas of appropriate brightness with the clone tool and duplicate the change in brightness across the wall due to the flash used when the photo was taken. Any abrupt change I cause really stands out and can be corrected immediately. This method is the most accurate, needs a relatively undamaged surface and can be time consuming
Because there are substantially undamaged areas of the wall, they can be used as the source for the healing brush to apply authentic texture/noise.
The next repair was done using the gradient tool, the circular gradient was selected because of the flash. Use the colour picker, set to 5 x 5 average to load the lightest area as foreground and the darkest as background. On a new layer guess/estimate where the centre of the flash was pointed, click on it and drag to where the darkest point was selected. You may have to try more than one centre point.
I usually put colour samplers at the darkest and lightest points and set the display for them to HSB, it makes it much easier to see how the result compares with the original. A levels, curves or hue/saturation layer can be attached to the gradient layer to trim it for the best match with the original, using the warp tool can also help.
I added noise at 2 pixels and Gaussian blurred it at 0.8 pixel
Gradient applied, with mask
(http://i151.photobucket.com/albums/s145/atholg/gradient_zpsbd0b9abb.jpg)
Zig Zag view
(http://i151.photobucket.com/albums/s145/atholg/gradientzigzag_zps56cab27b.jpg)
This method is best when there is a large amount of damage
In both cases the aim is create a repair which the average viewer will not identify as such. I often print my repairs at the correct size to see how near (or not) I am to that.
My first attempt at a tutorial so let me know if anything isn't clear
Athol
Hi Athol,
Thank you so much for this tutorial. I have a very busy day ahead of me so I won't be able to get to this until tomorrow (Friday). I'm intrigued by the zig-zag curve you used. I use a variety of curves patterns when I'm looking for dust and dirt on my photos. I plan to spend a lot of time walking through each step to make sure I understand them. If I have questions, I'll let you know.
Again, thanks for your help on this!
Judy
Those are interesting techniques Athol. I think I like the result from the zig zag curves better, but it's a little strange to work with :)
Tori
(http://img217.imageshack.us/img217/4080/millerl20164x6.jpg)
After all of the yellow color cast was eliminated, I found that both ladies have gray hair. I'm calling it finished & plan to move on to Challenge #2. However, I plan to lighten the darks - blacks
I found this challenge very time consuming & difficult, mainly because of the damage on the middle face.
(http://i1162.photobucket.com/albums/q538/bjtx/NON-Restore-Review-lighter-3_zpsf829453d.jpg)
Looking forward to more entries from the long-time members, as well as the newer volunteers.
btw: I will try to answer any questions about my processing, if there are any. :)
Athol, thanks for your latest tutorial.
Betty
Thanks Judy and Tori
In retrospect I think my description was probably back to front.
I was attempting to describe the two methods I have used to repair/replace backgrounds/flat surfaces ie clone and heal or gradient and I included an example of each.
For the first example using the clone/heal brushes, the curve allowed me to easily pick clone sources of appropriate brightness to continue the gradual change due to the use of flash. It also allowed me to see the typical noise in the undamaged parts of the original and to use those parts as source for the healing brush to achieve uniform noise/texture.
For the second example I used a gradient which gives a smooth transition between dark and light. The result has no texture and I added blurred noise as an example, using the curve allowed me see noise in the undamaged parts of the image and try to match it.
These methods work whether a zig zag curve is used or not, the zig zag curve is just a tool to allow you to look very critically at brightness in the image. The very small differences show up as changes in colour and is initially confusing. I have been using on all restorations for some years and have forgotten where the original suggestion came from.
As a suggestion, create the curve and use it to compare one or two of your repairs to the original. See what find.
Thanks Betty. At least you got to the end of it, I didn't. Only comment is that you have lost a bit of texture and I think the two dividers are parallel so the papers on the front one shouldn't be square and the base should match.
Athol
Athol, I see what you pointed out :up2: --- I'll make some corrections & then off to Challenge 2 :D
(http://i1305.photobucket.com/albums/s545/seekingoz/non-restore-challenge-working-2_zps3b4c43f6.jpg)
Hi Athol,
I finally tried the zig-zag method. I didn't get the same results you did but I can see where it could become a very important tool for repairs. I copied the wall results and blended them with my original repair. The changes look minor to me but I did put back a bit of the reflection on the white board and I got some texture in the walls. I didn't do any extra work on the middle face.
Great technique. I'll definitely be doing more with it. Thanks again.
Judy
Hi all
Since I haven't been up to attacking this monster, I've been keeping track of your progress, and it's phenomenal. It does seem to me, however, that all of this extra work should be compensated in some fair manner. My suggestion is to determine what you feel is a reasonable charge for your effort and put it on Margie's BILL!
Cheers and hugs
GK
Hello. I'm a brand new volunteer and I'm hoping there are easier images than this one to work on! Here is my WIP. I think the Lady In Red's face is beyond my abilities, but I may at least give it a shot. I think I can see what she looks like, I just haven't been able to turn her back into her old (natural-looking) self again.
Marty
Monterey County
(http://i1307.photobucket.com/albums/s582/haircut11/millerl20164x6_zpsd2c21e08.jpg)
Hi Marty, kudos for taking on one that tests your abilities as you will gain a lot of experience. Yes, these are all hard and they are ones close to being classed as non-restores. The reason these are being posted is so that we can all gain something by getting different ways to approach the hard ones. There will be no easy ones on the challenge, but once we get photos back from NY you will have ones that fit your comfort zone and should you run into problems, then we'll be here on the forum to help.
Hang in there!
Margie
This challenge has shown some amazing techniques to recover detail that seemed to be lost, thank you all for sharing these tips and posting your work!
I truly think think from what I have seen so far that this restore is very close to being done. What if we all post our final version of this restore within the next 7 days or so? Feel free to "borrow" parts of other volunteer's work to complete your own if need be. That way this challenge would be a real group effort. :)
Hannie
Marty, I'd like to borrow your whiteboard for my final version. Do you have a full-size copy available?
Tori
This is fun, we get to pick the good parts and merge them in our own restore!
Just like going shopping. :)
Hannie
Quote from: Tori803 on January 28, 2013, 06:18:45 PM
Marty, I'd like to borrow your whiteboard for my final version. Do you have a full-size copy available?
Tori
Hi Tori.
I'd be honored. Here is a link to the full size copy: http://s1307.beta.photobucket.com/user/haircut11/media/millerl20164x6_edit_zps12868a0c.jpg.html
I hope you can download it from there.
Marty
Thanks Marty! I was able to download it. I planned to finish up today, but was called in to work to fill in for someone, so it may be another day or two.
Hannie, this is a reduced size image with my final changes.
Hopefully, some of the parts will be useful.
Thanks for the challenge.
(http://i1162.photobucket.com/albums/q538/bjtx/non-restore-1-final-1000-px_zpsc11f6fc3.jpg)
editing notes fwiw: the blue channel was helpful for a guide. However, I used it sparingly on the middle face, since the channel contained so much grungy damage.
ORIGINAL:
(http://img217.imageshack.us/img217/4080/millerl20164x6.jpg)
I made a couple of changes. Final version (I hope)
(http://i1162.photobucket.com/albums/q538/bjtx/OPR/NON-Restore-GALLERY-2_zps407400e3.jpg)
Hannie, please let us know if we managed to complete a image which can be considered a restoration.
Thanks
Betty, I think the left cheek line (our left) of the middle woman is more accurate in your version than mine. I'll have to make an adjustment.
Tori and Betty, what beautiful results, you have done the impossible!
:wnw:
Hannie
Hannie, you're very kind. One thing I've learned from this challenge is the value of the input from other volunteers, especially on the problem areas!
Who would have thought it possible. :up2:
Great job Tori and Betty, would it have been regarded as a successful restoration?
Athol
Personally I don't feel my restoration is there yet. To be successful the faces have to be perfect. So I will keep tweaking...