Topics

Luminosity Masks


James Gray
 

I presume there are some strong and useful opinions on this list about all of the talk about luminosity masks as recommended by Tony Kuyper, Jimmie McIntyre, etc..  I have wondered for some time if I could take advantage of luminosity masking within the PPW.  I can see now how it could be done.  It is interesting that the Tony Kuyper and McIntyre way of producing luminosity masks is not the only way.  When I realized that you could use curves to produce luminosity masks I realized I have been using masks that could be called luminosity masks for 10 years.  I also use what are more accurately called chromosity masks built from the A and B channels of the LAB color mode.  In the past few days I have been trying to see what I could accomplish specifically doing luminosity masking.  I added 3 curves adjustment layers to a couple of images.  One had a shadows luminosity mask, another had midtones mask, and the third had a highlights luminosity mask.  I have some questions.  In several places I have read that luminosity masks work well because they are effectively feathered.  In testing in the last few days I have found that a curves adjustment layer with a luminosity mask does not work as well as I was hoping.  Detail seems to get lost when there is a significant amount of local texture or small detail.  However, if I put a fairly large blur on the mask local detail comes back in a way that usually looks far better than without the mask.  This is almost exactly what Dan has recommended in Professional Photoshop 5th ed and elsewhere.  Much of what I have been reading about luminosity masks elsewhere does not suggest blurring the mask or even seem to be aware of the option to blur.  It is pretty clear that blurring is not normally used.  The problem with blurring is that it tends to result in halos around edges.  However, the rocky slope of a mountain or the feathers on the wings of a bird looks much better with the blur applied.  Do other users of luminosity masks find blurring of the masks effective or useful?  I note that the actions I have to generate luminosity masks typically generate as many as 18 different masks.  In my testing in the last few days I was only using 3 masks.  Am I missing something.  Does having 18 different masks change the value of blurring in some way?  If so, how?  I would appreciate input, comments, or suggestions from anyone who has used or attempted to use luminosity masks.
James Gray


Massimo Calvani
 

I use Tony Kuyper's panel (V7) and I recommend it.
You can get TK Basic V6 panel here
You can also blur the mask, but usually not needed. At least this is my experience.

Massimo Calvani 



Michael Colby
 

Guy Gowan (guygowan.com) applies sharpening to luminance masks to deliver contrast that also increases clarity. sharpening tonal masks has some very good application and results. 


Dan Margulis
 

James Gray writes,

I presume there are some strong and useful opinions on this list about all of the talk about luminosity masks as recommended by Tony Kuyper, Jimmie McIntyre, etc..  I have wondered for some time if I could take advantage of luminosity masking within the PPW.  I can see now how it could be done.  It is interesting that the Tony Kuyper and McIntyre way of producing luminosity masks is not the only way.  When I realized that you could use curves to produce luminosity masks I realized I have been using masks that could be called luminosity masks for 10 years.  I also use what are more accurately called chromosity masks built from the A and B channels of the LAB color mode.  In the past few days I have been trying to see what I could accomplish specifically doing luminosity masking.  I added 3 curves adjustment layers to a couple of images.  One had a shadows luminosity mask, another had midtones mask, and the third had a highlights luminosity mask.  I have some questions..  In several places I have read that luminosity masks work well because they are effectively feathered.  In testing in the last few days I have found that a curves adjustment layer with a luminosity mask does not work as well as I was hoping.  Detail seems to get lost when there is a significant amount of local texture or small detail.  However, if I put a fairly large blur on the mask local detail comes back in a way that usually looks far better than without the mask.  This is almost exactly what Dan has recommended in Professional Photoshop 5th ed and elsewhere.  Much of what I have been reading about luminosity masks elsewhere does not suggest blurring the mask or even seem to be aware of the option to blur.  It is pretty clear that blurring is not normally used.  The problem with blurring is that it tends to result in halos around edges.  However, the rocky slope of a mountain or the feathers on the wings of a bird looks much better with the blur applied.  Do other users of luminosity masks find blurring of the masks effective or useful?  I note that the actions I have to generate luminosity masks typically generate as many as 18 different masks.  In my testing in the last few days I was only using 3 masks.  Am I missing something.  Does having 18 different masks change the value of blurring in some way?  If so, how? 

Masks based on channels (which are luminiosity masks unless we are specifically talking about the AB channels) have been in common use for some time and are a standard part of PPW in at least three contexts:

*The false profile scripts, where the objective is to reduce the difference between sunny and shady areas;
*The Color Boost scripts, when we sometimes find that the added color is more obnoxious in darker than in lighter areas.
*The H-K scripts, where they are used to prevent the script from plugging the shadows.

Blurring is needed when the difference in tonality (not color) between the two layers is larger. Therefore, the false profiles scripts need a big blur, the H-K could get by without one but the script uses a slight blur, and the Color Boost needs nothing at all.

It’s true that there is not much talk about when blurring is needed. I surmise that it’s because most people don’t make drastic tonal moves through these masks and/or because blurring the mask is rather counterintuitive if you haven’t seen a demonstration of it working. I myself discovered it by mistake, when I was trying to do something else to the mask and instead blurred it and was shocked by the improvement.

It’s rare for haloing to be an issue with the false profile actions. You’re more likely to see haloing when using Bigger Hammer.

How many masks are enough? Well, for me, if I find a typical sun/shade image and want to try a false profile script (change document to a lower-gamma [lighter] profile, then add a layer to multiply it, with a blank layer mask ready to fill with something) then my rule is, if the shadowy areas are more important than the lighter ones, use the RGB composite from the multiply layer as the mask, otherwise use the RGB composite from the background layer.

Now if it’s a particularly important image I would consider something else. Right off the bat there are eight channel masks available: the red, green, blue, and RGB on each of two layers. But I could also find a mask in LAB, or make some kind of combined mask by say, starting with an RGB channel and blending an AB into it in Linear Light mode. And if that doesn’t offer enough we consider the possibility of lightening the mask. For example, if I’m using a false profile script on a photo where I want a pronounced difference between light and dark areas, just not as strong as what the camera provided, I msy use a mask at 50% opacity or whatever. In summary, in my personal workflow, if the image isn’t out of the ordinary 18 mask choices are too many, but if it is then 18 is not enough.

Dan Margulis


Gerald Bakker
 

Here is an explanation for the blurring step.

Luminosity masking is often used on images with harsh contrast. The basic idea then is to "darken light image areas" and "lighten dark image areas". Sometimes only one of these is required, sometimes one stronger than the other, but the idea remains the same.

It is natural to do such corrections with some sort of luminosity mask. Make the image lighter but restrict the effect to darker areas, and vice versa. The concept of "darker areas" is crucial. A luminosity mask without blurring applied, will apply the effect on pixel level. Dark pixels will get lighter, light pixels will be left alone. This destroys local contrast, usually undesirable. Apply a large blur to the mask, and the mask discriminates light/dark areas as opposed to light/dark pixels. The price to pay is a risk of haloing.

Here is an interesting experiment. Open a low-contrast image in Photoshop. Add a curves adjustment layer and apply a rather strong S-curve. Assume you want to limit the effect to light areas only, so apply a copy of the RGB to the mask of the curves layer. This should more or less recover the shadows. We now have a version with better contrast in the lights. Duplicate the image, and on the duplicate apply a large blur to the mask. Compare the results, especially for the light areas.
 
I found there is hardly any difference, and where there is, the version without the blur looks slightly better (more contrast, as wished for). I think the conclusion is that blurring is only needed when the aim is to diminish contrast, not otherwise.
 
Gerald Bakker
http://geraldbakker.nl


James Gray
 

Dan,
Thanks for taking the time to reply.  I have actions that create 18 luminosity masks.  In evaluating the use of luminosity masks and what the 18 masks look like that the actions are creating, I am sure there is no reason those 18 masks have any bearing on the issue of blurring.  I have done very limited testing blurring or not blurring.  I should add that I am doing curves with a mask on an adjustment layer.  I have also loaded a mask as a selection before applying a curve to the green or red channel.  In most of my tests blurring the mask helped a lot.  In a few cases in made no difference.  So far I have not found an instance where the blurred mask was not as good as no blur.  So my results are not the same as Gerald's.  In some instances blurring did create ugly halos that had to be handled in some other way.  There are clearly other ways of utilizing luminosity masks besides with curves adjustment layers.  Some people do a lot careful painting often using soft light blending mode through a luminosity mask.  I have never really tested this approach.  I went to a seminar near where I live in which the speaker created multiple layers with different luminosity masks.  He applied the Camera Raw filter to each layer.  I have not tried that technique.  It seemed to work well for the speaker even if he had to spend a lot of time on each image.  I would presume that you are right that most people do not make drastic tonal moves through these masks so do not notice or even create the muddy look that I have seen with unblurred masks.  On a related note, I wish I could develop 25% of the intuition you have for what moves or steps will create good results.  I am particularly bad at judging in advance which of the hammers or similar steps will produce the best results.  That said, it was not surprising to me that blurring a mask could really help areas of an image with texture or lots of local detail like the breast of a bird.  Regarding halos, it is not hard to do selective blurring of a mask to prevent the creation of halos along important edges.  This is all very helpful.

James Gray


On Fri, Oct 11, 2019 at 10:02 PM dmargulis DMargulis@... [COLORTHEORY] <COLORTHEORY@...> wrote:
 

  I note that the actions I have to generate luminosity masks typically generate as many as 18 different masks.  In my testing in the last few days I was only using 3 masks.  Am I missing something.  Does having 18 different masks change the value of blurring in some way?  If so, how? 

Masks based on channels (which are luminiosity masks unless we are specifically talking about the AB channels) have been in common use for some time and are a standard part of PPW in at least three contexts:

*The false profile scripts, where the objective is to reduce the difference between sunny and shady areas;
*The Color Boost scripts, when we sometimes find that the added color is more obnoxious in darker than in lighter areas.
*The H-K scripts, where they are used to prevent the script from plugging the shadows.

Blurring is needed when the difference in tonality (not color) between the two layers is larger. Therefore, the false profiles scripts need a big blur, the H-K could get by without one but the script uses a slight blur, and the Color Boost needs nothing at all.

It’s true that there is not much talk about when blurring is needed. I surmise that it’s because most people don’t make drastic tonal moves through these masks and/or because blurring the mask is rather counterintuitive if you haven’t seen a demonstration of it working. I myself discovered it by mistake, when I was trying to do something else to the mask and instead blurred it and was shocked by the improvement.

It’s rare for haloing to be an issue with the false profile actions. You’re more likely to see haloing when using Bigger Hammer.

How many masks are enough? Well, for me, if I find a typical sun/shade image and want to try a false profile script (change document to a lower-gamma [lighter] profile, then add a layer to multiply it, with a blank layer mask ready to fill with something) then my rule is, if the shadowy areas are more important than the lighter ones, use the RGB composite from the multiply layer as the mask, otherwise use the RGB composite from the background layer.

Now if it’s a particularly important image I would consider something else. Right off the bat there are eight channel masks available: the red, green, blue, and RGB on each of two layers. But I could also find a mask in LAB, or make some kind of combined mask by say, starting with an RGB channel and blending an AB into it in Linear Light mode. And if that doesn’t offer enough we consider the possibility of lightening the mask. For example, if I’m using a false profile script on a photo where I want a pronounced difference between light and dark areas, just not as strong as what the camera provided, I msy use a mask at 50% opacity or whatever. In summary, in my personal workflow, if the image isn’t out of the ordinary 18 mask choices are too many, but if it is then 18 is not enough.

Dan Margulis


onelistdrs
 

There are several panels available. I have found, for me, that "Lumenzia" is very useful and easy to use/understand.
look here:

He has very helpful videos and listens to users. I started with ver.1 and all updates now to ver.7 have been no added cost. Can also see tips via Facebook/Lumenzia.

Doug Schafer


James Gray
 

Thanks for the link.  I looked at it.  Do you use the PPW as your workflow?  Are you able to able add the use of Lumenzia onto the PPW? If so, at what step?

James Gray

On Sun, Oct 13, 2019 at 7:43 PM ds-mail@... [COLORTHEORY] <COLORTHEORY@...> wrote:
 

There are several panels available. I have found, for me, that "Lumenzia" is very useful and easy to use/understand.
look here:

He has very helpful videos and listens to users. I started with ver.1 and all updates now to ver.7 have been no added cost. Can also see tips via Facebook/Lumenzia.

Doug Schafer


Dan Margulis
 


James Gray writes,

 On a related note, I wish I could develop 25% of the intuition you have for what moves or steps will create good results.  I am particularly bad at judging in advance which of the hammers or similar steps will produce the best results.  That said, it was not surprising to me that blurring a mask could really help areas of an image with texture or lots of local detail like the breast of a bird.

The intuition of experience is helpful in many contexts because without it one can spend a lot of time chasing wild geese and then having to revert back to the state before the chase began.

The PPW panel was designed to avoid this. One of its key, underrated features is that any script can be cancelled with a Command-Z. If you’re not sure whether Bigger Hammer, say, is going to work, try it and is it doesn’t then cancel it. If you’re not sure whether it is working then save it as a Variant and try something else to see if you can do better. Either way it doesn’t cost much time.

Where intuition would be more helpful is if you decide to mess with the Bigger Hammer options, like changing the blur radius or the channel being used for the mask. Same thing goes for the options that are available for many of the other scripts, like MMM. There, you can waste a serious amount of time by not having a good idea of what might work. The simple question of whether a hammer action will work doesn’t offer that unpleasant possibility.

Dan Margulis


Dan Margulis
 


Gerald Bakker writes,

Here is an interesting experiment. Open a low-contrast image in Photoshop.

I’m not sure what that means. If you are saying that it lacks a full tonal range then that should be corrected before proceeding with anything that’s going to require a mask. If you are saying low contrast epsecifically in the shadows and/or in the highlights, yet overall there is a good white point and black point, that’s fine.

Add a curves adjustment layer and apply a rather strong S-curve.

A curve in the shape of an S is relatively flat in the highlights and shadows and steeper in the midtone. I think that you are talking about an *inverted* S, where the shadows and highlights are stieepened and the midtone is flatter.

Assume you want to limit the effect to light areas only, so apply a copy of the RGB to the mask of the curves layer. This should more or less recover the shadows. We now have a version with better contrast in the lights. Duplicate the image, and on the duplicate apply a large blur to the mask. Compare the results, especially for the light areas.
 
I found there is hardly any difference, and where there is, the version without the blur looks slightly better (more contrast, as wished for). I think the conclusion is that blurring is only needed when the aim is to diminish contrast, not otherwise.

I do not claim a perfect knowledge of why blurring the mask works. If I am correct that you have two layers are using an inverted S-curve on the top one, however, then I question whether it is a valid test. A better test would be to apply a simple one-point curve to the bottom layer, lightening the midtone by about a third.

The difference between the two approaches is in damage control. My simple curve improves anything darker than a midtone and damages anything lighter, and damages the highlight particularly badly. By applying a channel-based layer mask to the top, untouched layer, the highlight problem is almost completely eliminated. There is still some damage in the quarter- to midtone range because part of the top layer is now being blocked out by the mask. My theory is that now blurring the mask disguises the remaining damage while permitting substantial improvement in the dark half of the image.

An inverted S-curve damages the midrange badly. A channel mask can’t hide that the way it can hide highlight damage. So there’s definitely going to be something odd-looking and probably disagreeable in the midrange. I’m not surprised that a blurred mask wouldn’t help.

Dan Margulis



James Gray
 

I normally use the snapshot feature of the history palette to save the hammer options and can easily compare them.  If any result looks like possible blending options then I save it as a variant.  It is true that the PPW panel is an excellent tool for avoiding a lot of time trying different things.  I rarely adjust radii or similar adjustments in any of the PPW steps.  It would probably not come as a surprise since I asked about masks that I tend to fiddle with masks on the MMM+CB step.  I do not know that I would call my time doing that a waste of time as I find it interesting and entertaining.  It is also hopefully improving my intuition for the various things I can do.  Photography and post processing costs me money rather than being a source of income.  So I have little reason to rush.

I will add that I have learned a lot about how the advocates of luminosity masking work in the last week or so.  Even though I had looked at the Lumenzia web pages a few months ago, the link that Doug Schafer posted was valuable.  Not only does does the owner Greg Benz provide a really simple panel for free.  He also provides some really informative videos for free.  He does suggest editing the masks.  Personally I think selective blurring might work better than the painting he demos, but that is my opinion.  The main take away for me is that the Benz or Kuyper approach to luminosity masking could be compatible with the PPW for high value images.

James Gray

On Tue, Oct 15, 2019 at 9:38 AM dmargulis DMargulis@... [COLORTHEORY] <COLORTHEORY@...> wrote:
 




The intuition of experience is helpful in many contexts because without it one can spend a lot of time chasing wild geese and then having to revert back to the state before the chase began.

The PPW panel was designed to avoid this. One of its key, underrated features is that any script can be cancelled with a Command-Z. If you’re not sure whether Bigger Hammer, say, is going to work, try it and is it doesn’t then cancel it. If you’re not sure whether it is working then save it as a Variant and try something else to see if you can do better. Either way it doesn’t cost much time.

Where intuition would be more helpful is if you decide to mess with the Bigger Hammer options, like changing the blur radius or the channel being used for the mask. Same thing goes for the options that are available for many of the other scripts, like MMM. There, you can waste a serious amount of time by not having a good idea of what might work. The simple question of whether a hammer action will work doesn’t offer that unpleasant possibility.

Dan Margulis


Beat C
 

> Dan Margulis wrote :
>The PPW panel was designed to avoid this. One of its key, underrated features is that any script can be cancelled with a Command-Z. If you’re not sure whether Bigger Hammer, say, is going to work, try it and
> is it doesn’t then cancel it. If you’re not sure whether it is working then save it as a Variant and try something else to see if you can do better. Either way it doesn’t cost much time.

I completely agree that the one-step UNDO is a very powerful in the PPW panel actions. It liberates one to freely experiment with a.o. the different Hammers, but also with different settings (obtained by holding the ‘option’ key when clicking a panel action). For me, with the ‘freely experimenting’ I am sharpening my intuition. I think intuition comes from experimentation which leads to a deeper understanding, which leads to good intuition. That’s probably why I always want to figure out the workings behind something, anything :-)

Beat



Gerald Bakker
 

>> GB: > Here is an interesting experiment. Open a low-contrast image in Photoshop.

>> DM: I’m not sure what that means. If you are saying that it lacks a full tonal range then that should be corrected before proceeding with anything that’s going to require a mask. If you are saying low contrast epsecifically in the shadows and/or in the highlights, yet overall there is a good white point and black point, that’s fine.

I meant an image that contains mostly midtones. One with a bell-shaped histogram. Black and white points are fine.  

>> GB: > Add a curves adjustment layer and apply a rather strong S-curve.

>> DM: A curve in the shape of an S is relatively flat in the highlights and shadows and steeper in the midtone. I think that you are talking about an *inverted* S, where the shadows and highlights are stieepened and the midtone is flatter.

No, I meant a real S-curve, not inverted. Such a curve increases midtone contrast, which is just what our image needs. It flattens highlights and darks, but that's not a real issue because there are not many of those anyway.

A luminosity mask has the effect of limiting the curve to the lighter parts of the image. The adjustment now stretches the lights while more or less retaining darks. There is still some gain of contrast. A blur of the mask now indeed reduces some of the local contrast, which usually makes the image look worse.

Whether this is a useful scenario, I don't know. But it shows my point that blurring is not always beneficial.

Gerald Bakker
http://geraldbakker.nl


Gerald Bakker
 

>> GB: > Here is an interesting experiment. Open a low-contrast image in Photoshop.

>> DM: I’m not sure what that means. If you are saying that it lacks a full tonal range then that should be corrected before proceeding with anything that’s going to require a mask. If you are saying low contrast epsecifically in the shadows and/or in the highlights, yet overall there is a good white point and black point, that’s fine.

I meant an image that contains mostly midtones. One with a bell-shaped histogram. Black and white points are fine.  

>> GB: > Add a curves adjustment layer and apply a rather strong S-curve.

>> DM: A curve in the shape of an S is relatively flat in the highlights and shadows and steeper in the midtone. I think that you are talking about an *inverted* S, where the shadows and highlights are stieepened and the midtone is flatter.

No, I meant a real S-curve, not inverted. Such a curve increases midtone contrast, which is just what our image needs. It flattens highlights and darks, but that's not a real issue because there are not many of those anyway.

A luminosity mask has the effect of limiting the curve to the lighter parts of the image. The adjustment now stretches the lights while more or less retaining darks. There is still some gain of contrast. A blur of the mask now indeed reduces some of the local contrast, which usually makes the image look worse.

Whether this is a useful scenario, I don't know. But it shows my point that blurring is not always beneficial.

Gerald Bakker


Rick Gordon
 

As a solution for isolating midtones, what about a mask calculated from the difference between a 50% gray channel and an auto-toned (if necessary to get full range) luminosity mask, inverted, and then auto-toned?

That leave elements closed to 50% most selected, while elements that are approaching 0% or 100% as least selected?

In other words:
  • Create a 50% gray channel

  • Run Calculations with Source 1 as L (or Gray, if in RGB or CMYK) and Source 2 as the 50% gray channel, in Difference mode.

  • Invert the calculated result channel, and then auto-tone that.
Rick Gordon

--------------------
On October 20, 2019 at 9:59:22 PM [-0700], Gc Bakker] wrote in an email entitled "[COLORTHEORY] Re: Luminosity Masks":
I meant an image that contains mostly midtones. One with a bell-shaped histogram. Black and white points are fine.  
___________________________________________
RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________
WWW: http://www.shelterpub.com


onelistdrs
 

Seems like a lot of work and things to remember when with a Luminosity program like TK or Lumenzia, it is as simple as try one of several midtone or color or sat or zones or range or zone pick or  diff or vibrance, etc. mask types (1 click to get each), pick the closest desired mask (may be with 1 click) and "tune" it with curves or levels, or paintbrush or?? ...and voila, mask is done to liking and can see mask visually and can see it's effect on a composite image used with any other layer adjustment such as curves, HSL, solid color, etc..

If no liked result, click a button to remove and try another....simple!

Sometimes quick results are easier if we "investigate" the image before looking for a mask.  Investigate to see what adjustments affect which parts of the image...this can be crucial to understanding the approach to fix an image and use a mask (or not) that will solve a problem area.

Blurred masks (none/little/a lot) can be useful depending on many factors; simple to try, see results, adjust amount, decide, apply or not.

Doug Schafer


Derick Miller
 

On Oct 13, 2019, at 21:58, James Gray
 

Thanks for the link.  I looked at it.  Do you use the PPW as your workflow?  Are you able to able add the use of Lumenzia onto the PPW? If so, at what step?

I have used Lumenzia, TKActions and PPW. 

PPW is in its own category. 

As for the other two: TKActions is substantially more powerful than Lumenzia. Their main purpose is for controlling how an effect is distributed  image, mostly by means of a mask. 

You can use them on anything, including results generated by PPW. 

The videos created by Sean Bagshaw to support TKActions are very good at giving you the basics of how to use the tools. Once you get what they do, you can figure out how you want to use them. 

Mostly they restrict what is affected. 

I continue to use TKActions. I don’t use Lumenzia. Lumenzia is not bad, but I use TKActions for things it cannot do. 



James Gray
 

[Doug Schafer]:  Seems like a lot of work and things to remember when with a Luminosity program like TK or Lumenzia, it is as simple as try one of several midtone or color or sat or zones or range or zone pick or  diff or vibrance, etc. mask types (1 click to get each), pick the closest desired mask (may be with 1 click) and "tune" it with curves or levels, or paintbrush or?? ...and voila, mask is done to liking and can see mask visually and can see it's effect on a composite image used with any other layer adjustment such as curves, HSL, solid color, etc..

[James Gray]:  I found this very confusing as I do not know what you are responding to.  I know a lot of people do not want long tails on their emails, but including part of what you are replying to for those of us who use email and not the web really could help.  Depending on what happens with the migration of the group it may be more important to include some of the words of the person you are replying to.  I think you might be referring to a reply by Rick Gordon.  I do appreciate your reference to tuning the luminosity mask.  That comment stands on its own.  It also helps me to understand how people use luminosity masks.  I have been tuning masks based on channels using curves for years.

 [Doug Schafer]:  Blurred masks (none/little/a lot) can be useful depending on many factors; simple to try, see results, adjust amount, decide, apply or not.

[James Gray):  Thanks.  Another helpful comment.


onelistdrs
 

...as a general addition, to the general topic of luminosity masks....

We can also use blend-if layer adjustment to act as a (luminosity or color) mask between layers.

Blend-if "mask" not seen but the effects are seen....however there is a .atn available to make the blend-if "mask" visible, if needed/desired.

Doug Schafer


James Gray
 

GB:  No, I meant a real S-curve, not inverted. Such a curve increases midtone contrast, which is just what our image needs. It flattens highlights and darks, but that's not a real issue because there are not many of those anyway. 

A luminosity mask has the effect of limiting the curve to the lighter parts of the image. The adjustment now stretches the lights while more or less retaining darks. There is still some gain of contrast. A blur of the mask now indeed reduces some of the local contrast, which usually makes the image look worse.

Whether this is a useful scenario, I don't know. But it shows my point that blurring is not always beneficial.

JG:  Gerald,  I do not understand.  I would like to see an example of an image where blurring a luminosity mask that has been applied to an S shape curve on an adjustment layer reduces local contrast.  Are you referring to a luminosity mask that is white where the S curve is close to horizontal?  Won't blurring of the mask improve local contrast where the S curve is steep.  Perhaps the way I have said this is not clear.  It seems to me that any luminosity mask created from an image with local detail will reflect that detail.  Blurring the mask will result in the detail being obscured in the mask, allowing the curve to bring out detail.

If the issues with Yahoo files is a problem, I would be interested in finding a way to share a file using another service.

James Gray