Colourise a black and white photo in Affinity Photo

[R C-R]

Everything you said is OK. But, explain to me how the scanner found those RGB values in grayscale image?

It is mostly because the scanned photo itself is not totally free of coloration. This could be due to aging of the emulsion, oils or other contaminates introduced by handling, reactions with cleaning agents used to prepare the photo, & so on. It is also possible there was a small amount of contamination on the glass of the scanner, too little to see except maybe under a very bright light, or there was a small amount of fluctuation in the color temperature of the scanner's light source during the scan.

 

Whatever the cause, the important thing here is the coloration has nothing to do with the colors of the photographed object. It is all from artifacts in the scanning process or the print being scanned.

 

That's why for best results, when scanning a "black & white" (actually grayscale) photo, the scanner should be set to grayscale mode.

[Petar Petrenko]

Just guessing I may say. Nothing exactly.

[Alfred]

The bottom line is: can you identify the colours in their clothing, and can you say whether the foliage in the background is green or brown? I can't.

[Petar Petrenko]

I can't also. But maybe "The new kids in the block" can do it?  ;)

[R C-R]

Just guessing I may say. Nothing exactly.

Try this: In AP add an HSL adjustment & turn up the saturation on the master all the way. This will intensify the color variations. Do you see anything that corresponds even vaguely to the colors you would expect the objects in the photo to have?

 

I can only do that with the tiny jpeg you uploaded, but even at that low resolution there is evidence of two reddish orange spots on the edges that probably are from oils deposited on the print during handling. If you do this with the original high resolution scan you should see more clearly what does & does not correlate to any colors of the objects in the photo vs. handling, aging of the photo, etc.

 

So yes, this is just guessing & nothing about it is exact, but that is the point. There is no way to extract accurate color information that wasn't recorded in the photo itself when it was taken. You might be able to reconstruct a vestigial amount of it if you have access to very detailed info about the camera, the lens, the film, & the lighting conditions, but that isn't likely & it is (as was mentioned earlier) an external reference, not anything you can get from the scan alone.

[Alfred]

Just guessing I may say. Nothing exactly.

 

What's your best guess about the colouring of the attached image before I desaturated it?

[Petar Petrenko]

What's your best guess about the colouring of the attached image before I desaturated it?

 

Please, reread this thread again if you have to. Don't just post to increase your number of posts.  :)

[mac_heibu]

Gray is gray us gray is gray. Of course these gray values have corresponding RGB values, because you can define every gray value as a RGB value. 50% gray, for example is RGB 128, 128, 128 – of something like that, because normally one has to take into account the color profile, which us used. One more problem: Different color tones may have the same gray tone – a big problem, when converting RGB into gray scale. Is 233, 233, 233 corresponding RGB value in a gray scale image cyan blue, or middle green? You can't re-calculate it. No way.

If you assign every gray scale tone or every corresponding RGB valua a specific RGB color, you get – the hell of a mess! Just because the well known 16,7 million of displayable RGB colors match 256 gray tones. That is the bitter truth.

[Petar Petrenko]

Thank you, mac_heibu  :wub:

[Alfred]

Please, reread this thread again if you have to. Don't just post to increase your number of posts.  :)

 

LOL. No, I don't have to re-read the thread, and I'm not just trying to increase my post count!

 

I was actually being a bit mischievous. The image I attached was one of three that I obtained by separating out the R, G and B channels in this image from Pixabay; that one is from the blue channel, so there was nothing but blue in it, but (as you will see from the explanation now offered by 'mac_heibu') there's no way the "new kids on the block" could have told you that.

[Petar Petrenko]

LOL. No, I don't have to re-read the thread, and I'm not just trying to increase my post count!

 

I was actually being a bit mischievous. The image I attached was one of three that I obtained by separating out the R, G and B channels in this image from Pixabay; that one is from the blue channel, so there was nothing but blue in it.

 

I don't think it is so easy or obvious. That is were the Affinity team comes into play.  :)

[Alfred]

I see that you replied while I was editing my post to add:

 

but (as you will see from the explanation now offered by 'mac_heibu') there's no way the "new kids on the block" could have told you that

 

I don't think the Affinity team can do any more about this than your "new kids on the block". :)

[R C-R]

Thank you, mac_heibu  :wub:

I'm not sure why you are thanking mac_heibu for saying pretty much the same thing we have been saying all along in a slightly different way.

 

A grayscale image has no color channels, no color information at all, just levels of gray. If it is an 8 bit image it can have up to 256 levels of gray. If it is a 16 bit image it can have up to 65536 levels, but there is no way to map those levels to 8 or 16 bit color channels to extract any color information from the grayscale image because there is no color information in it. Period.

[R C-R]

Hey Alfred,

 

You are looking very gray today. Feeling OK?  :lol:

[Petar Petrenko]

I was trying to finish this thread in which a lot of post are almoust repeating. Obviously, some of the users are lazy to read the thread from the start.  :D

[R C-R]

I was trying to finish this thread in which a lot of post are almoust repeating. Obviously, some of the users are lazy to read the thread from the start.  :D

 

What is obvious & to whom is a ... wait for it ... gray area. 

 

(Somebody had to say it. It might as well be me.)  :lol:

[Alfred]

Hey Alfred,

 

You are looking very gray today. Feeling OK?  :lol:

 

[Alfred]

What is obvious & to whom is a ... wait for it ... gray area. 

 

(Somebody had to say it. It might as well be me.)  :lol:

 

Well, I'm just glad it wasn't me! :D

[anon1]

https://www.youtube.com/watch?v=sdEyAwmasfo

german tutorial covering the procedure of coloring a photo in Affinity Photo (no PS tutorial)

 

the gist is to use at least 4 layers per part (for the hairs/ lips/ skin) and then mask them selectively to get a mix of colors for each part of the body 

 

cheers (and happy new year)

[anon1]

On 27.7.2016 at 8:33 PM, MBd said:

William, no Software in the world will ever Color a black and white image by itself (unless it is RAW)

I will admit....

https://www.reddit.com/r/InternetIsBeautiful/comments/4t8cuk/use_deep_learning_algorithms_to_add_color_to/

 

you can try for yourself 

http://demos.algorithmia.com/colorize-photos/

 

 

advice on how to do it in AP

https://forum.affinity.serif.com/index.php?/topic/19493-how-to-colorize-black-and-white-photos/&do=findComment&comment=231330

[dmstraker]

Landscape Pro has a simple way of letting you say 'this is a tree, this is a house' etc. Maybe this idea could be used for some intelligent colouration. AI should also be able to do such things. If it can recognized people, buildings and trees should be relatively simple.

[anon1]

7 minutes ago, dmstraker said:

should be relatively simple.

compared with this high quality work

https://forum.affinity.serif.com/index.php?/topic/42794-ap-colorization/

I doubt it is simple 

[R C-R]

28 minutes ago, MBd said:

On 7/27/2016 at 1:33 PM, MBd said:

William, no Software in the world will ever Color a black and white image by itself (unless it is RAW)

I will admit....

So called "deep learning" has the potential to do much more than colorize photos. If an artificial neural network can learn to do that, then they should be able to learn to do things like make & refine pixel selections or convert raster images to high quality vector tracings that are comparable to or possibly better than what humans can do using manual methods.

 

This could be a real game changer. Imagine a selection brush that is "smart" enough to select clothing or a person, or objects like trees or backgrounds, depending on what a user drags it across on a photo, or a "tracer" brush that can extract, vectorize, & group selected parts of an image.

 

The hard part (besides doing the training) is developing algorithms that can do all this locally, without sending anything to a remote processing service.