Color/Luminance blending modes versus the inherited ones?

[Alejandro]

There is in PaintShop a plain color blending mode and an inherited (legacy?) color blending mode. There also is a plain luminance blending mode and an inherited (legacy?) luminance blending mode.

In the manual (in the Spanish version of it at least) it says that the inherited blending modes are there to keep compatibility with previous versions of PaintShop. The results when using the supposedly newest blending modes are not the same as the ones that are obtained when using the inherited blending modes. I tend to like best and find more useful the results I get with the inherited blending modes (probably a matter of personal taste).

Does anybody know which are the technical differences between these modes (which is the difference between inherited luminance blending mode and luminance blending mode and between the inherited color blending mode and the color blending mode? What are they doing differently?

[LeviFiction]

Can't actually say.



Luminance and color are supposed to work like this

Luminance - Uses the the lightness from the top layer, and the hue and saturation from the bottom layer to make the new result.

Color - Uses the Hue and Saturation from the top layer, and the lightness from the bottom layer to make the new result.

The legacy ones do do this, and while Hue and Sauturation are always replaced properly they seem to come up with different answers for luminance. So sometimes the result is darker or lighter than it should be after mixing the colors. Though I can't tell you why this is.

[Zacabeb]

The legacy blend modes use the traditional Y'UV formula for calculating luminance and chrominance from R'G'B', where

Y' = 0.299 * R' + 0.581 * G' + 0.114 * B'
U = (B' - Y') * 0.492
V = (R' - Y') * 0.877

...which is inherited from analog color television, and also used in a variant for separating luminance and chrominance for JPEG, as well as a variant for standard definition digital video.

Note the prime symbol (') in R'G'B' and Y'.This indicates that the channels have been gamma encoded. CRTs in TVs and monitors have a non-linear relationship between the signal level and light output because of the way the electrons in the CRT are captured along the way to regulate light output, affecting the gamma curve - how bright midtones appear compared to shadows and highlights. To compensate for this, the signal is encoded with a complimentary gamma curve upon capture, often including some intentional undercompensation since the display would be realtively bright compared to the surroundings.

The standard used today on PCs for how gamma should be handled, sRGB, is based on this and assumes that the display has an output gamma of 2.2, and that pictures therefore should be encoded with a gamma of the inverse, about 0.45. In practice the gamma curve math is a bit more complicated. sRGB has become sort of a de facto standard and it's often assumed digital images are in sRGB unless their metadata states otherwise.

Now that we have flat panel displays, they actually mimic the behavior of CRTs, applying a gamma curve roughly similar to that of CRTs.

When recalculating saturation and brightness using the Y'UV formula, it doesn't take this gamma compensation into account meaning the results may not be visually optimal. But it's what the legacy modes do.

Jasc then introduced the new Luminance/Saturation/Hue/Color blend modes (in PSP8 I think) which reverses the gamma encoding of the R'G'B' values to bring them back to linear space, applies the respective math, and then reapplies the sRGB gamma curve to the result. I don't know exactly what formula is used and how 'smart' it is with respect to colorimetry and so on. But the idea was to produce a more accurate result which would also better mimic what the mudhut was doing with their 'professional' software. Ironically, the mudhut had changed their math similarly but skipped on having legacy modes, rendering peoples existing projects using those blend modes butchered. Jasc were smarter than that.

To summarise, the legacy modes blend directly in gamma encoded space, and the new modes blend in linear space with unknown math.

[Alejandro]

LeviFiction, thank you for your input, I really appreciate it.

Zacabeb, yours was a very thorough and complete technical explanation. Thanks a lot. I had to read your explanation several times before I could more or less grasp the main concepts you explained (though it is nice to have also been able to look at the formulas you brought up and try to derive some additional conclusions from them), but I think I finally managed to end up understanding at least which and why are the differences between the new and legacy blending modes. Thanks again.