But is it any worse than the extremely rare contemporary colour footage, which is itself extremely unreliable?
And with the hardware we don't know with any great accuracy what the real colours were. The Solomon schemes certainly included bright colours.
The second set down is quite plausible. The top one could be way off. Some recent research (at Bovington?) suggests that all Fosters-built tanks were painted in their commercial Brunswick green. Noting that sets 2, 3 and 4 are the same tank in the same place - with the same group of soldiers in sets 3 and 4.
Colours we do know: the local mud, uniform fabric, grey undershirts and web equipment (surviving examples), boots, skin tones. Only set 2 comes anywhere close. The others are way off.
But I ask again if they are any worse than contemporary colour photography. Noting that we also argue over the colours in WW2 photos 20-odd years later.
I am reminded of the devices that were on sale when colour TVs came out, these claimed to 'convert' your B&W TV, in reality they were a filter you stuck over the screen, that was 'blue' at the top and 'green' at the bottom, so on the right scene they worked.
Please, please don't post these "colourised" pictures. At some point they will be reposted more often than the original black and white photos and noone will be able to find the originals anymore.
It is impossible to reconstruct the true colours from black and white data. While, in certain cases, it is possible to make educated guesses based on the values in a b/w photograph, the filter curve of the film and the known possibilities of the paints which could habe been used at the time, the process these programs use is far too simple and only aims to provide a visual plausible result, not a realistic one. Using them in this context just helps to spread confusion amongst all who don't know the origin of the false colours.
Please, please don't post these "colourised" pictures. At some point they will be reposted more often than the original black and white photos and noone will be able to find the originals anymore.
It is impossible to reconstruct the true colours from black and white data. While, in certain cases, it is possible to make educated guesses based on the values in a b/w photograph, the filter curve of the film and the known possibilities of the paints which could habe been used at the time, the process these programs use is far too simple and only aims to provide a visual plausible result, not a realistic one. Using them in this context just helps to spread confusion amongst all who don't know the origin of the false colours.
Thorsten
I doubt that there will be much confusion as there are not very many original colour photo's from WW1.
Most of the photo's I have tried so far have not yielded good results.
Hopefully the software will be developed and may get better with more reliable results.
I still feel that it is a useful tool to help identity paints, to be used alongside known availability of paints etc.. When the software does work it can produce
interesting results.
From what I have read there does seem to be a great deal of uncertainty or at least disagreement about what colours were used.
Maybe the results are far away from original. But I think this depends on the point of view.
I guess it was more than 15 years ago when I discussed with IT-professionals (all of them Mensa me´mbers!) about developing a Software which can colorize black-white-photos.
My thought was that b-w-pictures are more then only black an white - they have variations of grey. So if I have a software which I can tell that parts of the b-w-pic are individual colours, like horizonblue, earthbrown or whatever, then the software could generate the other colours from that information.
The IT-pros told me that this would be impossible also in the far future, because some colours will give the same grey also when they are very different. For example Charlie Chaplins Great Dictator: there is an existing colour movie from the making of the film, which shows that the trousers of the SA are red in reality.
So from that point of view: the results are great. And how about the future? If the Software can learn more, then ... maybe .... one day .....?
No it is indeed impossible. And it has a very simple reason. In a computer, a colour is often represented by three numbers, one each for red, green and blue. These numbers give you the intensity of the three colour channels, together they control the final colour of a pixel. If you dive a bit deeper into computer graphics, you will see that you can also change the representation of a colour from red, green and blue to hue, saturation and value. Together, combinations of these three values cn be used to build every colour from the rgb-system, just using three different values.
And this is the key: a colour is only identified completely by knowing THREE values. A black and white photograph gives you only one number, which is exactly the "value" parameter from the second system mentioned above, and multiplied by the filter curve of the film. There is just no information about the other two values you would need to identify a colour. And this is why each of the above colourised images is just pure fantasy.
The reason why the pictures look as good as they do is probably because the program has a face recognition algorithm, assigninh skin tones to the respective pixels, and a preference for having a blue sky in the upper half.
The result does not help you in any way, and I again hope that we won't be flooded with this kind of pseudo-historical images. They take away the real value of the originals.
In some period photos, Mother appears to be completely unpainted with a metallic sheen. Little Willie was certainly grey, and Mother was certainly later repainted. In her scrap photos she is in a darkish shade with a lighter horizontal stripe.
Vehicles, as with steam locomotives, were often painted grey on roll-out for their PR photos as "engineer's grey" came out especially well in B/W photos. But then repainted in service livery. I do not imagine the Brunswick Green suggestion applied to Mother initially, if at all.
I agree that colourising without accurate references is nothing but a fool's errand. As noted, a shade of grey could be one of several colours. We must accept that we just don't know with any real accuracy what the original colors were because of the lack of standardisation and codification that did not really start until 15-20 years later. Occasionally some period paint will be discovered, such as the Weald Foundation's restoration of their pair of FT's. As they removed paint they had it analysed by a University and they believe that their sand/red-brown scheme was the original factory finish on both vehicles. But that was just one of several schemes used on FTs. They also discovered that the interior was 2-tone grey and white. But opportunities for such original discovery are getting fewer and farther between as too many period vehicles have been "restored" without such care and analysis. One wonders what Saumur discovered in restoring their CA1. Some original period paint mixing recipes are known, but unless you make the same mix from the same ingredients you are unlikely to achieve the same colour with model paints. The oil base and drying agent added tints to the colour as well as the white lead base and the powdered earth/mineral pigment.
I looked into Buntfarbenanstrich for a Beute MkIV a while back. The only surviving references are some original stahlhelm, which exhibited a wide variety of shades of all the colours. Brown from chocolate to burgundy, green from grass to brunswick and sand from cream to yellow. Which supports the assertion that colours varied according to local mixing - in a way that colourisation can never cope with.
I think that as a matter of good practice and the avoidance of any misunderstanding, if you colourise a photo you should add a "colourised" comment on the face of the picture itself.
Well ... I don't agree, because it is again a point of view.
The first question to answer is: are the colours unrealistic? Can you see any colour which is not anly far away from the original, but totaly unrealistic? Is there a colour like pink for the uniforms? No. Is there neon-green where brown or yellow should be? Take a look at the Pictures and the answer must be no.
Second question to answer is: Will it be impossible in the furture to generate realistic, true coulors near to the originals? This does not depend on the point of view. This depends on how software will work in the future. Coming back on what I said: based on the answers 15 years ago I was told that also what we see here as a result of software generated colours should only maybe in 30 years possible. I think that, as the develpoment of software and especially AI progresses, we will see nearly true colours.
No, it is impossible, and everyone who knows a bit how the processes work will agree here. A small illustration of the problem: I chose three numbers between 0 and 100 in secret. I sum them up and tell you only the result. The task for you is to calculate which three numbers I had at the beginning, which is impossible, as you just lack information. Any combination of three numbers which sum up to the same value are as good as your guess, and this is exactly the same problem as you face with trying to work out colour from bw images. The problem cannot be solved, this is a mathematical fact.
Software can improve to give you more plausible results based on experience, but it will never be able to bring back information which is lost. Colour is lost when exposing a bw film.
This does not depend on view but can be proven scientifically. What does depend on your point of view is how you like the results above, but this does not say anything about the realism.
I think there is a misconception about what is "realistic". You seem to assume that visually pleasing and colours which are in agreement with your expectations are enough to call them realistic. If that's the case, fine, but I would not call them realistic, as I'd also require them to be based on some evidence, for example colour samples, knowledge from written sources or so on, or simply because they come from a real colour photograph.
And that's the problem I see here. People who don't know how the colour was created will assume that it was done to the best knowledge we have, because it looks good enough for them. This undermines real research, as it creates wrong impressions and, at the first glance, makes real research superfluous (because seemingly a simple computer program can produce 'realistic ' colours).
If the software produces some kind of colour which looks ok but is based on nothing else, the result for me is worthless. And I hope that I was able to explain why it just can not deliver any hints of real colours. Not even a bit. It is pure fantasy.
I know that we all want to know what the WW1 colours were. In this information age we don't like to be denied information. But apart from a very few surviving pieces and the odd paint mix recipe we simply don't know except in the most general terms. If the information was ever noted, specified or captured it is now lost. Computerised guesswork - and today it is little more than that - does not advance the quest in any objective way. Anything it produces can only be subjective, making it no better than any other subjective opinion or stylised period painting. This is not to say that the future might not hold something different, although I struggle to see how. But then if I could see how, I'd be making my fortune seeing it!
The factory model of the MkVIII at Bovington is reputed to be the only known original sample of the brown used on late-war British tanks. But even that has to be tempered by saying that it is the shade used by one factory (Beardmores?) in 1918 and cannot be taken as representative of any other brown used by anyone else. It is also 101 years old and has certainly weathered in the light. It is now quite a greenish brown, unlike the more milk chocolatey colour often depicted. So is it "right"?
I'm not sure that we help ourselves by getting so wrapped around the axle about knowing the precise colours and peddling our own particular opinions. In this debate there is room for multiple opinions and I don't think that any one of us has the high ground. But I think we all know that if we took a group of people dressed in surviving original uniforms and equipment and recreated that photo scene in a muddy Somme Valley field it would look nothing like any of those colourised photos. Because we don't know what is right, by definition we don't know what is wrong either. At least, not what is subtly wrong rather than howlingly wrong. Come WW2 we have codified colours, known colour values, surviving paint chips, samples and painted artifacts. And we still argue endlessly about the subtlety of colours.
And then we come to aging of the original prints, different film stock, camera settings, over/under exposure of the negative and the printed image, different scanning values, different graphics cards, different monitor types and settings and our own eyesight. All of which affect our perception of an image we see on screen and some of which affect the digitisation pre-colourisation. Far too many variables. In these days of precision and perceived perfection we find it hard to recall that it was not always like that, and that technology can't always improve it. Like the old lady whose relatives had a favourite photo digitised only for her to ask why they hadn't removed some object her dog was hiding behind so that she could see it.
