Bit Depth for Layer Masks


Today’s Question: Just read something today I’d never heard of and wondering if you have any input on this. I came across this searching for potential improved methods for hair masking. He states:

“…there’s a hard-wired inconsistency in Photoshop. Selections are always 8 bits. Once you load a 16-bit channel mask as a selection it becomes an 8-bit selection, and therefore virtually eliminating the 16 bits depth of the 16-bit masks.”

He did go on to say he’d not seen significant differences in the masks but I’ve come to trust you over the years and wonder if you had any thoughts.

Tim’s Quick Answer: Selections and layer masks in Adobe Photoshop are indeed 8-bit per channel even for 16-bit per channel images, but I don’t consider this to be of any concern at all. In other words, I don’t feel it will make any difference for a photographic image to attempt to create a mask in 16-bit that will only be converted to 8-bit per channel once the mask is used to apply a targeted adjustment to an image.

More Detail: The primary “risk” of working with your images in the 8-bit per channel mode is posterization, or the loss of smooth gradations of tone and color. For a typical color image in the 8-bit per channel mode, there are up to almost 16.8 million possible color values available. As a result, posterization is not a high risk unless you apply very strong adjustments to an image.

For black and white (grayscale) images the risk of working in 8-bit per channel mode is more significant. That is because an 8-bit black and white image can only have a maximum of 256 shades of gray, as compared to a maximum of 65,536 possible shades of gray for a 16-bit black and white image.

However, the concerns of posterization don’t really extend to layer masks. To begin with, very strong adjustments are generally necessary before any degree posterization will be visible in a color even at 8-bit per channel. If you’re working in 16-bit per channel mode for either black and white or especially a color image, the risk of posterization are not especially high. It would take very strong adjustments to create visible posterization, and I would suggest that an image requiring very strong adjustments in the first place probably wasn’t the best photo.

When it comes to layer masks (and by extension selections), the bit depth is not anywhere near as important. Even a somewhat posterized layer mask would still be mitigating the effect of even a strong adjustment across the image. It would require significant posterization on a layer mask combined with a rather strong adjustment being constrained by that layer mask in order to create any visible artifacts in the image.

Keep in mind that a somewhat typical layer mask might involve defining a “stencil” for the image that is divided between white and black, with some gradation between those areas to blend the targeted adjustment. Even with a more sophisticated layer mask, the fact that the mask only consists of up to 256 shades of gray would not lead to any concerns in terms of artifacts of other image quality issues.

So, I would say there is no benefit to creating the form of a layer mask in 16-bit per channel mode considering that mask will be converted to 8-bit per channel mode as soon as it is loaded as a selection or used as a layer mask in Photoshop.