Accurate perceived monochrome rendering of colour images

[Richardgb]

Advertisement (gone after registration)

This is a topic directed at those with theoretical and practical experience of rendering colour images into accurate perceived monochrome. The easy (lazy?) way to do it is simply to desaturate the image, but this results in some very odd renderings of, for instance, skin tones.

Going a little deeper, I found one explanation in a good introductory video here: https://www.youtube.com/watch?v=gGZ7UV1I0xU. The video also demonstrates the 'good' (or, rather, 'better') ways to convert compared to simple desaturation.

Some programs (like Affinity Photo, demonstrated in the video) allow for individual adjustment of primary (RGB) and secondary (YMC) colours, but what values should any adjustments be set to? (Moving all by an equal amount would be equivalent to desaturation). The video suggests 30% R 59%G 11% B which seems to be going in the right direction since the eye is most sensitive to 'green' region of the spectrum, least sensitive to 'blue', and red somewhere between the two.

Fortunately the video does contain a few clues about where further information may be found. However, by doing so it becomes clear that these RGB values* were originally derived for the NTSC television system so may not apply to rendering from, say, an image in the sRGB or Adobe RGB colour spaces. (In fairness, the presenter does say where her RGB values came from but does not point out the possible shortcomings).

Help is at hand, though, in Lindbloom's (2017) RGB Working Space Information page, where alternatives to the NTSC figures are given for, amongst others, sRGB and Adobe RGB. See http://www.brucelindbloom.com/index.html?WorkingSpaceInfo.html

The figures he gives are (to 4 decimal places):

0.2973R 0.6273G 0.0752B for Adobe RGB (1998)
0.2127R 0.7152G 0.0722B for sRGB images.
0.2988R 0.5868G 0.1143B for NTSC RGB images (essentially the same as those in the video above)

As he notes, '... it should also be obvious that the real RGB weightings depend upon the color system in use. So the "standard" weightings [for NTSC] are incorrect for other RGB systems like sRGB or Adobe RGB (1998).'

Finally, I am aware that these values are just a starting point and that further tweaking of the colour sliders may be desirable to get the right 'look' for a particular image.

Please share your experiences of colour to monochrome conversion.

*See also C. Saravanan, "Color Image to Grayscale Image Conversion," 2010 Second International Conference on Computer Engineering and Applications, 2010, pp. 196-199, doi: 10.1109/ICCEA.2010.192.

Edited March 1, 2022 by Richardgb
typo

[jaapv]

The most accurate way to convert to monochrome is to move into LAB. Basically RGB colour spaces are not the best choice for their integration of colour and luminance.  For the method, see "The Canyon Conundrum" by Dan Margulis pp 110-114, the method is a bit too convoluted to discuss in a forum. Having said that, the standard conversion tools in LR 22 are pretty good

[wda]

I have no problem using Lightroom. BUT I rely on my judgment of what is a good monochromatic rendering. A lot of people, particularly males, have some degree of colour perception issues.

If you think you suffer in some way, from recognizing any colour channel,  I suggest you try using black and white conversion presets.

[zeitz]

16 hours ago, Richardgb said:

perceived

When I took philosophy, the principle was "Quidquid precipitor, precipitor secundum modum precipientis".  I am sure I am not spelling the Latin words right, but it translates as "whatever is perceived, is perceived according to the mode of the perceiver".  I have no idea what an accurate perceived monochrome is because my eyes are not monochrome.  And as WDA says, even perception of color is different for every individual.  My perception of color changed after cataract surgery.  So if we take a Leica M Monochrome, it never has color because it has no Bayer filter and is accurate to the degree of the linearity of the sensor.  Just don't use a B&W filter.  But even B&W is affected by the color temperature of the light (time of day, weather).  All digital sensors are monochrome, only achieving color through the use of a Bayer filter.  The Foveon sensor doesn't use a Bayer filter, but it captures the three colors at different layers of the sensor, more analogous to color film.

Also like WDA says regarding presets, I normally use Silver Efex for B&W conversion, tweaking its controls.  I used to use Channel Mixer in Photoshop, but Silver Efex essentially put an end to that.  I tried the LAB color space method referenced by Jaap, but I never got on with it.

[Richardgb]

6 hours ago, zeitz said:

When I took philosophy, the principle was "Quidquid precipitor, precipitor secundum modum precipientis".  I am sure I am not spelling the Latin words right, but it translates as "whatever is perceived, is perceived according to the mode of the perceiver".  I have no idea what an accurate perceived monochrome is because my eyes are not monochrome.  And as WDA says, even perception of color is different for every individual.  My perception of color changed after cataract surgery.  So if we take a Leica M Monochrome, it never has color because it has no Bayer filter and is accurate to the degree of the linearity of the sensor.  Just don't use a B&W filter.  But even B&W is affected by the color temperature of the light (time of day, weather).  All digital sensors are monochrome, only achieving color through the use of a Bayer filter.  The Foveon sensor doesn't use a Bayer filter, but it captures the three colors at different layers of the sensor, more analogous to color film.

Also like WDA says regarding presets, I normally use Silver Efex for B&W conversion, tweaking its controls.  I used to use Channel Mixer in Photoshop, but Silver Efex essentially put an end to that.  I tried the LAB color space method referenced by Jaap, but I never got on with it.

Thanks. 'Perceived' in the sense I am using it means how colours of equal luminance would be perceived by the eye (assuming normal vision) if they were represented in shades of grey. As the video I referred to shows, if colours of equal luminance are simply desaturated, they all look the same (unrealistic) shade of grey whereas if they are rendered using the percentages given, they seem more 'natural'. (As the video and other sources say, the problem arose in the early days of colour television when colur receivers were relatively rare and the signal had to be rendered on both colour- as well as black and white TVs, hence the NTSC reference).

Until the late-19th century all photographs were black and white, and even today photographers, particularly those of a certain age, would have used black and white film and become accustomed to how colours were rendered and how this may be affected by colour filters over the lens. Before the late-1980s, too, newspaper images were reproduced in black and white unless there was a special occasion (colour imagery was reserved for magazines), and no-one complained the images were unrealistic on such grounds. In other words, humans have the ability to accept black and white rendering of colours.

Note, though, that I am not trying to emulate this or that black and white film. However, I presume the creators of Nik's Silver Efex took into account how certain black and white films would 'perceive' colours (i.e. according to their different spectral sensitivities, notwithstanding that they were all panchromatic), rather than the eye, and built these in to the software's presets.

[250swb]

Use Silver Efex Pro, you can play with the channels in there, apply filters, presets, whatever you want to do, it isn't as complicated as the OP is making it out to be.

[Jeff S]

Advertisement (gone after registration)

Even after conversion, editing to taste, etc, the print process further involves variables related to paper choice and profile, inkset, printer capabilities (including print head and driver/RIP).  I use ImagePrint to fine tune print rendering after adjustments in LR (and sometimes Photoshop). Tonalities and tonal relationships can be influenced at every stage. And even after all that, rendering can be significantly affected by cover glass, display lighting, etc.  

Perception? At the end of the day, results are dependent on user decision making....at every stage from shot to final display.

Jeff

Edited March 2, 2022 by Jeff S

[frame-it]

1 hour ago, Richardgb said:

represented in shades of grey.

perhaps something like :

 

https://www.rnib.org.uk/nb-online/life-in-black-and-white

Edited March 2, 2022 by frame-it

[jaapv]

It is interesting to know that, when Leica designed the Monochrom (M9), they tweaked the coatings on the sensor to mimic the tonal scale of Delta 100. Which demonstrates that there is not one "correct" tonal scale. As Jeff points out, the end result depends on many factors which are mostly choices and judgments by the creator of the image.

[pop]

8 hours ago, jaapv said:

Which demonstrates that there is not one "correct" tonal scale

How could there be? Is there any evidence that there exists a world-wide consensus about a particular purple splotch being more or less bright than a greenish one, or equally bright? And in both candle light and sunlight at noon? When even the eyes of a single person can render the same color differently ?

[jaapv]

Exactly - but that is what the OP presupposes.