Why RGB and Lab read brightness differently
In RGB, each channel, Red, Green and Blue, carries both color and tonal information tangled together. Photoshop's composite luminosity read (what you get from Ctrl-clicking the RGB thumbnail) is a weighted blend across all three, which works well in practice but is still built from channels that were never purely about brightness to begin with.
Lab mode splits the two apart entirely. The L channel holds nothing but lightness data, no color information at all, while the A and B channels carry the color separately. Using the L channel as your starting point means you're building a luminosity mask from data that was never mixed with hue or saturation in the first place, instead of extracting brightness back out of channels where it was always mixed in.
The workflow, step by step
1. Work on a duplicate document
Converting a document's mode is a real color-space remap, not a free operation, and repeated round-trips between RGB and Lab can introduce small cumulative shifts. Duplicate the image first (Image → Duplicate) and do the mask-building on that copy, so your original RGB working file never actually leaves RGB.
2. Convert the duplicate to Lab and isolate L
Image → Mode → Lab Color. In the Channels panel, the L channel is now your clean lightness data. Duplicate it to create your base "Lights 1" mask.
3. Build narrower zones with Calculations, not repeated selections
To narrow toward the brightest zones, use Image → Calculations with the Multiply blend mode to intersect Lights 1 with itself, producing Lights 2, then repeat on Lights 2 to get Lights 3, and so on. For Darks, duplicate and invert Lights 1 (Ctrl/Cmd+I) and run the same Calculations process from there. Use Calculations rather than loading and reselecting, since it keeps the math in full 16-bit precision instead of round-tripping through an 8-bit selection.
4. Build midtones separately if you need them
A midtones mask isn't a simple intersection of Lights and Darks. Duplicate the L channel again, desaturate and apply a Curves adjustment shaped to peak in the middle of the tonal range, then blend that through Calculations the same way, since midtones need their own dedicated construction to land cleanly in the center without a bias toward one side.
5. Bring the finished channels back to your RGB document
Once your masks are built, copy the finished channel(s) from the Lab duplicate's Channels panel over to your original RGB document (drag the channel thumbnail into the RGB document window, or select all and copy/paste into a new channel there). Close the Lab duplicate. Your working file stays in RGB throughout the actual edit; only the mask-building step ever touched Lab mode.
Is this worth it over the RGB composite method?
For a fast, everyday luminosity mask, the RGB composite Ctrl-click is still the right default, it's faster and close enough for most corrections. The Lab-mode route earns its extra steps when you need the cleanest possible separation between tone and color, typically in color-grading work where you don't want a hue shift accidentally leaking into what's supposed to be a pure brightness selection.
Quick answers to a few related questions
Does converting to Lab mode and back lose image quality?
A single round-trip on a duplicate you discard afterward is not a practical concern. The risk is repeated, unnecessary conversions on your actual working file, which is exactly why building the masks on a duplicate and only bringing back the finished channels avoids the issue.
Can I just apply adjustments directly while the document is in Lab mode?
You can, but some filters and adjustments behave differently or aren't available in Lab, and it changes how your whole document is organized. Building masks on a duplicate and returning to RGB for the actual edit sidesteps that entirely.
Why use Calculations instead of just loading each mask as a selection and reselecting?
Loading a selection and rebuilding from it works, but it round-trips through 8-bit selection data at each step. Calculations keeps the intersecting math in the document's native bit depth throughout, which matters more the more zones you chain together.
EN
IT