Something like this? I guess they've used a

semi-automatical approach: automatic segmentation

and human suggestions what colors should be

used and which segments should be logically

connected.

Beer bottles should be brown, wine bottles green.

No AI program can do that yet.

http://www.popsci.com/article/gadgets/civil-war-color

http://dynamichrome.com/

http://www.interesting6.com/color-brings-old-photos-to-life-39-pics/

http://www.techeblog.com/index.php/tech-gadget/24-historical-photos-that-have-been-colorized-and-brought-back-to-life

Regards,

Joachim

You have no enough information to do it from single image (without "white balance info").

http://edition.cnn.com/2015/02/26/us/blue-black-white-gold-dress/

Hello Richard,

If you want to address the question fundamentally, you must study Colorimetry and color space management. That can not be easily summarized here.

However, if you are merely looking for a quick hack and a simple way to solve a problem at hand, take a look at "absolute color spaces" such as "Lab". But more importantly don't overlook what Dmytro mentioned. White balance correction is a very crucial step. Photographers often use a standard "color chart" as a reference for camera color calibration. You might want to take a look at these (just few examples):

http://www.csie.ntu.edu.tw/~fuh/personal/Calibration-BasedAutoWhiteBalance.pdf

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.70.6709&rep=rep1&type=pdf

https://hal.inria.fr/hal-00845866/document

http://www.red.pe.org.pl/articles/2014/9/49.pdf

At the end you must keep in mind, the choice of color space and white balance correction technique very much depends on the context of your problem.

Hope this helps :)

Cheers,

Saeed.

Thanks you all very much! I think with all this advices I have a good head start trying to solve my particular problem.