B&W Conversion 3 – Luminosity

Today we’ll look at the 3rd of three B&W conversion techniques, Luminosity.

sharps lane cades cove

Cades Cove, Great Smoky Mountain NP

Misty Dawn Scene


This technique produces what I call a “luminosity” B&W.

  • It can be done with any editing program that uses layers and has a Luminosity blending mode. The method (using Photoshop Elements) is shown in the following screen shot:
    • Make a copy of the original image
    • Place a white layer beneath it
    • Change the blending mode of the copy to Luminosity
    • You’re done

The result derives from how Luminosity blending works.

  • Luminosity takes its hue & saturation values from the 2nd layer (the white layer below the luminosity blend layer) and its luminosity values from the top layer.
  • Since the 2nd layer’s H & S value are zero (neutral) the result, as shown on the left, is pure black and white (gray tones) whose lightness values match those of the original image.
  • Voila – we’ve extracted the luminance channel.

For the 4th & final time, here’s the comparison of all three conversion techniques

2015-08-29_7-37-21


It’s unfortunate that image editing software programs offer desaturation and grayscale, but not luminosity – unfortunate because, of the three, it’s the best. There is an easy way to make a luminosity-based B&W without going through the PS routine described above. Tune in tomorrow,


Tomorrow – Luminosity Tips


Enter your email address to follow this blog and receive notifications of new posts by email.

7 thoughts on “B&W Conversion 3 – Luminosity

  1. While I – generally speaking – do agree that luminosity is a powerful conversion method, I find the message of luminosity-based conversion being “the best” a bit blunt. I’m too lazy to create an example, but I’m sure that I could come up with a very colorful, expressing picture that, if converted with the luminosity method, would yield a uniform grayscale. To me, the conversion algorithm depends on the original image – try a few different ones every time and compare the results. But possibly start with luminosity-based conversion, yes 😉

    • Just like “always” and “never”, “best” may not be totally accurate. Maybe we can agree on “one-of-the-best”. Having said that, your “but I’m sure…” may be a similar stretch. I’m inclined to doubt that you can come up with a uniform grayscale via luminosity conversion. Show me and the other viewers, please….

      • Now you just allocated my coming weekend for me – hopefully it’ll be a rainy one 🙂 OK, maybe it turns out to be a stretch, yes. But I will certainly put some effort into it – since a grayscale conversion turns 16-odd million combinations into 256 there must be more than one the other way round…

        • That left me restless 🙂 While digging a bit deeper, I realized that the nomenclature might be a bit confusing. Here (https://dl.dropboxusercontent.com/u/10393116/Permanent/lumi.zip) is a sample of what you call “grayscale” but GIMP and other web references seem to call luminosity-conversion to grayscale. The attached example shows both a desaturated grayscale and a “grayscale”[Knepley et al] or “luminosity”[GIMP et al] version of a colour image where the colours have been determined by a simple Perl hack creating a CSV file of RGB values all yielding a grayscale of 128 (assuming a grayscale/luminosity conversion of 0.21*R+0.71*G+0.07*B. There are slightly differing opinions out there what the best formula is, but this is the one used by GIMP, amongst others.

          I can’t post pictures here, hence I just link to the ZIP file. Maybe you’d like to create a post out of it?

          While this doesn’t prove my earlier hypothesis, I am trying to understand what your Photoshop-recipe actually does in mathematical terms. I suspect it refers to the L-part of the HSL model but am not quite sure at this point. Unfortunately I have neither Photoshop nor Windows, so I’ll have the additional challenge of trying to recreate the recipe in the GIMP. Well, the weekend is coming, I still have some time to provide another sample which will hopefully also create havoc in Photoshop 🙂

        • You mention grayscale here. The original point of contention is on what I called luminosity conversion. I explicitly described exactly what I meant for each of the three techniques and provided formulae for desaturation & grayscale. My “luminosity” was defined by showing the technique in Photoshop.

          You objected my use of “best” related to luminosity and said “… I’m sure that I could come up with a very colorful, expressing picture that, if converted with the luminosity method, would yield a uniform grayscale…”. If you choose to do what you said, please do it via luminosity conversion and not some variant of the grayscale technique – otherwise it’s apples & oranges.

          Good luck…

          • I thought that’s what I said — trying to figure the mathematics behind your luminosity method, Ie your apple, so that I can also create apples. Unfortunately I don’t have an apple tree so I was hoping someone might hint about how to create an apple without the tree? What I found so far is pointers that Photoshop’s luminosity seems to be the “lightness” value of HSL. But I’m not sure, those are forum-guesses. Well, I have to keep searching I suspect…

            • Hajo, the answer (it’s simply the HSL model’s L-value) to “the mathematics” is clearly stated right in this post where I wrote in reference to my conversion example:

              “…Luminosity takes its hue & saturation values from the 2nd layer (the white layer below the luminosity blend layer) and its luminosity values from the top layer.
              Since the 2nd layer’s H & S value are zero (neutral) the result, as shown on the left, is pure black and white (gray tones) whose lightness values match those of the original image.
              …”

              So, you can make an apple without the tree by extracting, pixel by pixel, the HSL’s L-value while completely ignoring the hue & saturation components. SMOP (simple matter of programming). 😉

              It’s interesting that without even knowing how my “best” luminosity conversion is done, you were sure you could create a uniform gray conversion with it given the proper image. “…I’m sure that I could come up with a very colorful, expressing picture that, if converted with the luminosity method, would yield a uniform grayscale…”. That can’t be done unless you use a color image where every pixel has the same L-value (this would be your “proper image”, but I don’t know how “…colorful, expressing…” it would be. 🙂

              Interesting conversation; keeps us on our toes. Thanks for your interest & have a nice weekend.

Comments are closed.