I got this photo to work on but I haven't a clue on how to get the faces cleaned up. Anyone have any ideas?

Hi QuietJan.

This is a tough one.

There is very little that can be done with this.

Please could you rescan the image in colour and re post the image in colour then we may be able to help more

Ken.

Jan

Ken is right - did u scan it?
Is it a negative or a print?
Is the picture really that bad?

The face detail is completely lost - so no information could be recovered.

If the oriinal picture isn't that bad - then try scanning at different levels - That is one overexposed to compensate for the shadows and one underexposed for the highlights.

Then combine in photoshop.

Look for High Dynamic Range (HDR) tutorials on the web.

Tmax100
London UK

No I didn't scan this. It was sent to me in email just the way it is. Is there nothing I can do with it this way? Maybe I can ask the person who owns the photo to rescan it using your suggestions.

Hi QuietJan!

Welcome to RetouchPRO!!! :pleased:

I have to agree with Ken and Tmax100... I tried minimizing the jpg artifacts and zoom in just to see if there was something...anything that could be recovered .... no go ... :sad:

IMO, depending on the conditions of the original, a rescanning could be the only possibility...

Was looking for a glimer of detail also... NOPE, couldn't see much!
Would love to see your solution QuietJan

this one is not totally impossible, but it's certainly missing some key bits, particularly the gentleman's eyes. those can be built, if you're really devoted to fixing this, but bear in mind that reconstructions are a TON more difficult and always full of uncertainty, since you're mostly guessing.

i can see the end of his nose and it does appear he's wearing horn-rimmed glasses. i can also make out the mouth and ear. but the eyes and eyebrows are shot. you can make out pretty closely where the eyes shld be, though.

the way i found these features was to convert the image to 8 bit rgb rather than b&w, then duplicated the background and used various blend modes to bring out some of the detail. no single blend mode did it all, though 'screen' in Paint Shop Pro worked pretty well and dodge blends in Photoshop worked fairly well.

the baby isnt in quite so bad a shape, but it's still going to take some work. you can make out the eyes on the child.

all in all, though, you're looking at a VERY tough restore/reconstruction.

this one is not totally impossible, but it's certainly missing some key bits, particularly the gentleman's eyes. those can be built, if you're really devoted to fixing this, but bear in mind that reconstructions are a TON more difficult and always full of uncertainty, since you're mostly guessing.

i can see the end of his nose and it does appear he's wearing horn-rimmed glasses. i can also make out the mouth and ear. but the eyes and eyebrows are shot. you can make out pretty closely where the eyes shld be, though.

the way i found these features was to convert the image to 8 bit rgb rather than b&w, then duplicated the background and used various blend modes to bring out some of the detail. no single blend mode did it all, though 'screen' in Paint Shop Pro worked pretty well and dodge blends in Photoshop worked fairly well.

the baby isnt in quite so bad a shape, but it's still going to take some work. you can make out the eyes on the child.

all in all, though, you're looking at a VERY tough restore/reconstruction.

Wow Craig... I'm always amazed by the power of Photoshop ... but to rebuild a face (never seen) in a credible/acceptable way having only the tip of a nose, the shape of a ear, the probable position of where-the-eyes-should-be to work on, and no other picture of the same subject to use for grafting, is really beyond me .... so, my total admiration for anyone capable of doing it!!! :bow:

I probably "could hang on" for a while to the left ear, but that's all.
I am afraid Flora makes good points.

but to rebuild a face (never seen) in a credible/acceptable way well, that's the trick, isnt it :) it's really a forensic sort of thing. you take known values and work from there, the outline of a face, a jaw, the tip of a nose, the depression at the side of a face, the distortion of a partial pair of glasses and so on and build from there. that it might be credible or acceptable...well, you and i both remember 'tub girls'. that one beat me up pretty badly :)

ok, this is extremely crude, but i've outlined some of the major things i can detect, if extremely faint. put that all together and i think something could be done, especially if another image of the gentleman exists somewhere.

and, as a tip, it's not a bad idea on a really bad image like this, to make one layer with some of these 'forensic' marks on it as a reference for working.

the really odd part of this image is the glass refraction. it seems to not be quite in the right place, but it may be due to the specs being down from his eyes a bit. but i'm convinced it is glass and therefore glasses.

well,...it's really a forensic sort of thing. ...That was actually the point I was trying to make ... :wink:
The reasons behind forensic are 'way' more pressing and of completely different nature than the ones behind restoration... also, the methods used are very different..

As far as I know, Forensic works with an incredible database of possible features to be used with the aim/in the hope that somewhere, someone could recognize the subject portrayed in the end result of their manipulations ...

Restoration aims to 'bring back' a picture to the time it was taken in the most natural-looking and as-close-as-possible-to-the-original way possible... where the the subject/s must always be recognizable by anyone who knows/knew them... (I should have added 100% recognizable to credible/acceptable...)

...well, you and i both remember 'tub girls'. that one beat me up pretty badly :)..I surely do.. had a lot of fun and you know I don't scare easily from very difficult/nearly impossible jobs.. :pleased: ...(for those interested, here (http://www.retouchpro.com/forums/photo-restoration/10423-ghostly-image.html) is the link to that Thread)... but, in my opinion, that picture had far more details than this one...

I'm fairly new to photoshop and for me something like this would be a MASSIVE effort. It really all dependes on how well the clinet wants it done. You will have to rebuild this photo using parts of other photos, eyes, ears, mouth, stuff like that. So if you can get other photos of these people that would be a start. I did the eyes and ear kind of quickly of the gentelman but I don't know what he looks like so it's difficult for me. Looks to me he has a mustache and a cigar. This photo will take alot of your time if you really want to do it well, many hours if your new like me. But it's worth it, for me anyway, just for the learning experiance. Good Luck, Shay.

hi flora,

well, when i used the word 'forensic', to me, that sort of implies that you arent going to get a totally accurate reconstruction, much like in police work; you're just trying to get as close as possible with a bit of deductive reasoning. i was also using it somewhat figuratively.

and all that, of course, would also depend on the client's needs and wants. some would simply say, dont bother; too much time, too much effort and too much cost for too little return. but there might also be someone willing to give it a shot and be happy with whatever could be done. thus, i thought maybe something could be done.

besides, i'm a bit compulsive ;)

ok, i dont have the time to try a full reconstruction/restore on this, but i'll post what i have so far and perhaps someone else can work with it, preferably someone who's a better artist than me. i think i may have the eyes set too close.

Jan - have you found out whether your client is able to rescan it or add another photo of the man and/or baby?

This could be totally wrong, but after working with lightening the dark blobs away, I thought that I saw this. Anyway, it still lacks most of the nose, but thought it might be "in the direction" of the original image.

Sorry for taking so long in getting back here. I had another person's photo I was working on to be ready before Christmas.

No I haven't heard back from her yet. Not surprising with the holidays and such. I think you may be onto something CJ. I've downloaded your photo to take a closer look at it. Thanks for the tip. I'll see what I can come up with using your method.

You have all given me some really good ideas and tips to work with. Yes I'm pretty committed to fixing this photo (maybe I should be committed...lol). I just hate to be defeated by a photo, just stubborn I guess. Thank you all for your input and I'll let you know how it ends up. Oh and yes the man is wearing wire rimmed glasses, I was told that by the owner of the photo.

Merry Christmas to all!

Happy holidays all :)

I applaud you for trying, however, I doubt if a logical end result can be obtained.
Knowing where the eyes and other facial features are will not give you an identifiable image, especially not one that would have value in reuniting ones attachment to someone that currently or once existed. Outline detail is only part of the equation. Highlights, shadows and character lines are just as important to an image to figure out the depth of given features that make us all unique. For example, if you were to construct an image having only placement of features the possibilities are endless unless you have other images of the same individual at various poses (brow, cheekbone, nose, character lines, etc). All of this information is necessary to associate the image to the individual.
Even if you had half of the face to work with and built upon this, you’d still be guessing. Most people are seldom recognizable if you duplicate a right or left side and combine since faces are not symmetrical. The little oddities are what matter.

To site an example; I have a program that does police composites, I’m sure many of you are familiar with it. In order for this program to be of value, the individual that describes the image must describe depth as well as dimensions in order to discern nose shape, eyes, cheeks etc., identification is otherwise hardly possible especially with only a frontal view. I guess my bottom line is, without more complete information, what’s the point?

Still, bear in mind that there can be a lot of information hidden in the channels. I always scan in RGB (16-bit, these days) for that reason. Always worth a try.