|
October 29, 2009, 12:48 PM
|
|
I think it may well be worth while getting a conversation going here on this topic. I can contribute and hope others will do the same. Accepting that the main difficulty is dealing with the colour changes due to degradation of the dyestuffs used to create the original colours then it may be useful to explore what methods are available and what results they produce. Some examples of degraded slides would be helpful and I am willing to make some available if needed.
|
| |
|
October 29, 2009, 02:14 PM
|
|
Yes, indeed. I was looking through my color slides last night to find one to tackle. If examples are posted, they should be TIFF's or XCF's but not JPG. The one I was planning to work on is cropped to only 5.2MB.
Can we post TIFF's to 23hq, and if so, can they be so large? I upgraded my membership, so maybe the limit (if any) doesn't apply.
|
| |
|
|
October 29, 2009, 02:39 PM
|
|
I tried uploading the TIFF to 23hq. TIFF uploads are not allowed. I suggest we post a jpg of the original here (for comparison), then post the TIFF to a Web server and post the link here. Ironically, my Web host is doing an upgrade today and tomorrow and has temporarily disabled FTP.
|
| |
|
|
October 29, 2009, 03:32 PM
|
|
I agree with you, jpg for comparison and, as most of mine are likely to be xcf will that be OK on a webpage?
|
| |
|
|
October 29, 2009, 03:55 PM
|
|
I'm not sure what you mean by "on a Webpage." I assume you mean that you would FTP the file to some space you own on the Web, then post the link on 23hq so people can download the file.
XCF is larger than TIFF, even with just the "Background" image and nothing else. For example, the photo I had in mind is 5.3MB for TIFF, 8.7MB for XCF. Other than size, they are equally good.
|
| |
|
|
October 29, 2009, 06:10 PM
|
|
|
|
| |
|
|
October 29, 2009, 08:58 PM
|
|
That would be excellent. How do I get the image to work on, please? Out of interest, about how old is the image and what film is it on?
|
| |
|
|
October 29, 2009, 08:59 PM
|
|
Perhaps I should read the words under the image.
|
| |
|
|
October 29, 2009, 10:48 PM
|
|
Re the film: The photo is from my father's collection. It was taken in the late 1940's. The negative was cut from a strip and trimmed so there are no markings identifying the film stock. It is 6 x 6 cm (2-1/4 square) format, probably 120 film and probably Kodachrome.
I just took another look at the raw file in ufraw. I discovered that the orange color cast can be removed in the White balance tab simply by lowering the temperature and increasing Green. So perhaps the color problem is not the result of film dyes fading with age, but rather of faulty white balance preset. And if so, we should choose a different photo.
|
| |
|
|
October 30, 2009, 07:15 AM
|
|
Thanks for the background information and I do not agree with you about the colour balance in UFRaw. Of course. you can modify colours in UFRaw but the colour cast in the image is a well known phenomenon. When you copy your positives does your camera make a jpeg at the same time? If it does, have a look at that image.
|
| |
|
|
October 30, 2009, 03:07 PM
|
|
My camera can shoot raw+jpeg but I always shoot in raw only to conserve my battery. Having raw files only is a minor inconvenience if you use the ufraw plugin. I almost always do some adjustment in ufraw before pressing the Gimp button. But maybe t here is an advantage to raw+jpeg that I do not appreciate.
Anyway, let's get on with this one (I will try to "restore" it later today) and then move on to one of yours.
|
| |
|
|
October 30, 2009, 05:30 PM
|
|
I have put 2 different attempts to recover the slide in the photo group. I prefer to use the EZA method because more detail is preserved and there is plenty of opportunity to use levels and curves if desired. I now have to search to find a suitable slide of mine.
|
| |
|
|
October 30, 2009, 09:12 PM
|
|
|
|
| |
|
|
October 31, 2009, 04:01 PM
|
|
Before restoring the challenge image, the chromatic aberration was removed using a CA correction plugin. Restoration was then completed using the procedure described in Episode 58 of MTG. The black point was chosen amongst the foliage in the left foreground and the white point was selected in a part of the sea where I wanted some detail. The image has been posted to the photogroup.
|
| |
|
|
October 31, 2009, 07:12 PM
|
|
Regarding challenge2. I can see detail in the foreground trees, and detail is generally good in shadows. But I am still troubled by the bluish color cast.
It seems to me that the main problem in color photo restoration is undoing the effects of ageing dyes. I associate the blue (or possibly magenta) cast with Agfa color slides in the old days, which turned magenta after a just few years on the shelf.
Assuming a color cast in challenge2 is the result of ageing, we should be able reverse its effect using Curves on the individual channels -- and no other tools! Since the color channels are R, G, and B, but the dyes are C, M, and Y, and since I don't understand color theory at a very high level, I can't predict exactly what the curves should look like. However by playing with curves, we should be able to get a good result.
I posted another "restoration" of challenge2 here
http://www.23hq.com/photogroup/5034048/photo/5056242
The only thing I did was manipulate the Levels ("curves for dummies") for the R, G, and B channels. Nothing else was done, not even sharpening. The result is not perfect, but I think it has dealt with some of the issues of color ageing.
|
| |
|
|
October 31, 2009, 08:04 PM
|
|
This whole question about colour and colour casts raises an argument that can go on for a very long time and brings one into the realm of the psychology of colour and how the eye and brain inter-react. We have a useful tool in the colour picker which should not lie so, if there is a colour cast, then the colour picker should expose it. According to my measurements, the image I posted does not have a blue cast although the eye might suggest there is one. I know there is no colour aberration because I removed it so where does the apparent blue come from?
|
| |
|
|
October 31, 2009, 08:28 PM
|
|
I don't mean to minimize the importance of subjectivity when looking at a photograph. I might actually prefer the bluer version because of the mood it creates, etc. But read on....
Try your color picker experiment again. This time, check "Sample merged" and maximize the radius (300). Then sample any area on the photograph. You will not be able to find one in which the blue channel is not dominant. At least I could not.
I looked up chromatic aberration in Wikipedia. It is the failure of a lens to focus all colors at the same point. Sometimes it manifests itself as color fringing around edges. Anyway, it is a localized phenomenon. Aging in dyes is a global phenomenon. It affects the entire photograph, like my old magenta Agfa slides. CA plugin won't fix it.
I found this interesting discussion on the Web about how dyes age. Look for the post by Shawn Kearney
http://photo.net/bboard/q-and-a-fetch-msg?msg_id=006myF
|
| |
|
|
October 31, 2009, 08:38 PM
|
|
You are correct about chromatic aberration being localised as colour fringing around edges and there are lots of them in the picture. They are there because the close up lenses that I have are not very good quality and are therefore not achromatic. If I remember correctly there is a very interesting and informative example of how to remove colour casts in the old GIMP bible Grokking the GIMP. This depends on the use of the colour picker to measure the cast and then uses curves to correct it. I will try to find it and, if I can, give it a whirl.
|
| |
|
|
October 31, 2009, 08:43 PM
|
|
Well I am about to try the Restore plugin on challenge2. According to the description, "it reverses the aging process." You used it on challenge1. Let's see what it can do with challenge2 (I will post the result).
|
| |
|
|
October 31, 2009, 09:03 PM
|
|
I posted the result of using Restore plugin with default parameters, and also one with "Degree of restoration" set to the maximum.
I think Restore does a really good job. I would go for the maximum. I'm really interested to know how it works. I will attempt to read the code.
|
| |
|
|
November 01, 2009, 10:47 AM
|
|
I agree that the Restore plugin is capable of producing very interesting results especially when the images are relatively simple. However, there is a marked loss of detail, in the Restored images in Challenge2 and these losses are exemplified in the histogram as well as the images themselves.
|
| |
|
|
November 01, 2009, 02:49 PM
|
|
I see what you mean. It it does seem to have lost quite a bit of detail and there isn't much of a histogram left to work with.
In my opinion, you can get just as pleasing a result by adjusting the R, G, and B Levels. This can be done in a single step. You retain plenty of detail, and the resulting histogram is spikey, but not as bad as when using Restore.
If Gimp had Adjustment Layers, such Levels adjustments could be done without losing information. The solution Rolf proposes in episode 124 (New from visible) addresses workflow, but not information loss, since a New from visible layer effectively becomes your new "Background," hiding the layers below. This defeats one of the main advantages of the layer stack, namely, the ability to go back and make adjustments in lower layers to immediately observe their effect on the final image.
Ufraw doesn't have a Levels tool, and I don't think Rawtherapee has one either. However Rawtherapee has Colour shift, which is probably designed for color slide restoration. We should try it.
|
| |
|
|
November 01, 2009, 04:04 PM
|
|
The more I get involved with using GIMP the more I keep muttering 'if only'! I accept what you have to say about colour correcting using levels or some equivalent it's just that, in this digital age, I think that one should be able to measure the RGB value of the cast and correct by dead reckoning. The method described in Grokking the GIMP attempts to do this but, like all these clever examples, uses an ideal image to show how to do it. I'll have a look at what I have and see if I can find something to try the technique out on.
|
| |
|
|
November 01, 2009, 05:10 PM
|
|
I was thinking the same thing. Why not a Gimp tool that tells you "This photo has a blue (or whatever) cast." The Colors -> Info tools don't include anything as straightforward as this. If you select RGB in the Histogram tool, you will see the blue curve dominates the others in the midtones and highlights. If I hadn't seen the photo, but was only shown the histogram, I don't know if I'd be able to predict that the photo had a blue cast.
But you don't have to be Sherlock Holmes to see that challenge2 has a blue cast. You look at it, and there it is. It helps to know that with age, certain films take on a blue cast and others take on a pink cast. This is most likely one of the former.
I read the chapter on color casts in the Grokking the GIMP book. He doesn't directly address color casts arising from ageing dyes in old color slides. If he did, I think the discussion would be shorter and simpler. Just move from Blue to Yellow, or from Magenta to Green (or something like that).
I tried Colour shift in Rawtherapee. You can get a very nice result by moving the Blue - Yellow slider to the right, and fiddling some with Green - Magenta. Then save as 16-bit TIFF and open in Gimp. Voila, the histogram is smooth as can be.
My only complaint about Rawtherapee is the single-window interface. It's a nightmare on a small screen. It doesn't allow the panels to be torn off and docked, like Gimp and other complex tool environments. I'll try to get used to it because it has a lot of the functionality I'm looking for in a Raw processor.
|
| |
|
|
November 01, 2009, 06:12 PM
|
|
Not even the great Photoshop has such a tool. To deal with colour casts first identify using the colour picker and so on. I think blue is a particularly difficult colour to deal with as is red and if you were a subscriber to the MTG Forum you would have come across an experiment I set up to try to show the red problem. I think Rolf may well pick it up and use it in one of his videos.
I can appreciate your problem with Rawtherapee on a small screen and I am glad I invested in a 21" monitor when I needed to buy a new one. I am not a control freak but, coming from a long experience in the darkroom, I do like to be able to control as much as I can. I intend to revisit Episode 58 and really try to work out what is going on because, in theory, it could well be the way to go.
|
| |
|
|
November 01, 2009, 06:55 PM
|
|
I tried your MTG experiment and left a comment.
An LCD monitor for photo work is out of my price range -- at least $1200 US.
|
| |
|
|
November 01, 2009, 09:20 PM
|
|
Saw your comment, thanks you certainly had an interesting result. I don't claim that my monitor is anything special just a bog standard LCD cost about £120 UK.
|
| |
|
|
November 02, 2009, 05:13 PM
|
|
Challenge3 is available at http://www.littletank.org/challenge/ and is a cityscape of Frankfurt in Germany. The slide, Kodachrome ll, is about 30 years old and was shot from the top floor of a tall hotel. For what ever reason the hotel people refused to allow the window to be opened so it was shot through the glass. Thus, in addition to the degradation of the dyes and the inevitable chromatic aberration there is the not very clean glass to be dealt with - enjoy.
|
| |
|
|
November 02, 2009, 06:37 PM
|
|
The chromatic aberration is very noticeable on the left and right sides. Not so much in the center. I guess I should install the CA plugin for this one.
Restore plugin at 50% gives nice colors while preserving detail, e.g., in the trees. At 100% it overdoes the restoration. Again, you can get just as good results fiddling with Colour shift in Rawtherapee.
The black spots look like dust spots on the lens (I almost said "sensor"), or maybe on the slide, rather than the window.
Without having specific criteria for what constitutes a good restoration, these challenges seem overly subjective. Except for the blemishes, chromatic aberration, and possible removal of color cast, any other "improvements" are in the eye of the beholder. Or are they? I need to read that book.
|
| |
|
|
November 02, 2009, 09:13 PM
|
|
I am glad you have already found time to have a go and, yes, CA is strongest at the edges, which is often the case. The black spots are dirt on the glass mount enclosing the slide and I was not really concerned about the as they can be cloned away without any real problem. I will do my restoration effort using a revised version of Episode 58 and then we can compare notes.
You raise an interesting point which, oddly enough, I had raised on the Forum earlier today in a slightly different form. I refered to the various ways available for restoration and asked how does one determine which is the most successful. Is there some sort of test or is it simply a case of 'I like that one best?'
|
| |
|
|
November 03, 2009, 09:12 AM
|
|
I forgot to mention - Rawtherapee has the facility to deal with CA.
|
| |
|
|
November 03, 2009, 03:30 PM
|
|
My restored image has been posted.
Do you want to continue with this theme to explore colour slide restoration or is something else in the restoration field you would like to have a crack at?
|
| |
|
|
November 12, 2009, 10:25 AM
|
|
I have recently bought a job lot of colour slides all of which are quite old and are commercially made copies sold to tourists and so on and, as you would expect, the colours have deteriorated significantly. There are slides of places in the UK, Italy and Israel and, before I do anything with them, I will try to find out whether the firms which made and sold these copies are still in existence. I would not want to fall foul of copyright laws and regulations. Once everything has been sorted out to my satisfaction you would be welcome to come and play, if you want to.
|
| |
|
|
November 12, 2009, 03:24 PM
|
|
That sounds like fun. I would love to see them.
|
| |
|
|
November 24, 2009, 11:40 AM
|
|
|
|
| |
|
|
November 24, 2009, 05:42 PM
|
|
Thank you for bringing these to my notice they really are astonishing but no amount of high resolution scanning would be of use unless the originals were in absolutely tip top condition. I would be surprised if there has not been a considerable amount of restoration and enhancement to arrive at the quality of those images.
|
| |
|
|
November 26, 2009, 05:29 PM
|
|
I have posted 35 images, each showing a before and after image, which are the result of a first approximation try at restoration. They have all been treated, as near as possible, exactly the same so there will be variations in the results and I would like to know if you see any general trends between similar sorts of images. It is hoped to be able to use your comments in the next stage of the development of the process.
http://www.littletank.org/restore/restore/
Thanks in advance for your help.
|
| |
|
|
November 27, 2009, 02:07 AM
|
|
I'd like to know what you did. You took care of the red cast but a green cast remains.
The first photo seems to have been done differently from the others. With the possible exception of the purple/violet reflections in the floor, the colors are completely believable.
I couldn't get nearly as good results using levels or curves. Better to use layers...
If too much red (ff0000), we need to push the compliment, which is cyan (00ffff). Adding a cyan layer in Hue mode with opacity < 100% does a decent job, though I'm not sure why. I'll post my version of #4 to the "Photo Restoration" group.
OK, how did you do it?
|
| |
|
|
November 27, 2009, 08:31 AM
|
|
I started from the basis that certain colour reversible films suffer from degradation and loss in the cyan layer hence the red colouration. You were part of the way there in your thinking so have a look at:-
http://forum.meetthegimp.org/index.php/topic,746.msg6029.html#msg6029
see if that helps and remember, this is only a first approximation.
|
| |
