Recovering highlights with shadow/highlights

rutt · April 5, 2005

It seems that a lot of us know how to use image->adjustments->shadow/highlights to recover shadow detail, but it is equally useful for recovering highlight detail. I've noticed quite a few messages over the last month or so about making two raw conversions and then recovering highlights by using a layer mask and painting out the blown details to allow the darker image to show through.

There is a global (no PS brushes or selections) alternative to this that can result in very natural looking results and which requires no skill or time with the wacom tablet. Essentially:

Adjust exposures and shadow in ACR so that neither end of the histogram is clipped.
Convert the raw into 16 bit mode
Apply shadow/highlight with low shadow amount and large highlight amount
Restore contrast with curves

Here is an example. I started out with this image:

I wanted to restore some of the blown detail in the clouds. I followed my recipe above and ended up with this:

Here is the shadow/highlight dialog I used to restore the shadow/highlight details:

And here are the LAB curves I used afterward to fine tune the contrast and improve saturation:

bfjr · April 5, 2005

thanks Rutt

Use all your tips and have been working all my images in LAB.
Question how do I emulate hightlight/shadow in PS7 or do I have to upgrade?

rutt · April 5, 2005

bfjr wrote:

thanks Rutt
Use all your tips and have been working all my images in LAB.
Question how do I emulate hightlight/shadow in PS7 or do I have to upgrade?

Shadow/highlight is a CS thing. To get the same effect in earler versions, you have to do some plate blending and very odd things with curves. It's much harder. There is some stuff about it on this thread: http://www.dgrin.com/showthread.php?t=5922&highlight=dynamic+range

I'll bet CS upgrades will be really cheap on ebay now that CS II is out. Probably worth it for ACR and shadow/highlight alone.

Andy · April 5, 2005

excellent result, rutt!

ian408 · April 5, 2005

If I had to pick one thing I love about your contributions, it'd have to be
these little pearls of wisdom. Well written, informative and very timely.

Thanks rutt!

Ian

rutt · April 5, 2005

andy wrote:

excellent result, rutt!

I think this will work almost anytime that you can recover the details in ACR. Sometimes they really are blown. Sometimes if you lower exposure enough to get back some details they will be magenta only. When this happens you can't just use this trick without doing something to "colorize" that magenta. But in this case you couldn't just use a layer mask with two differnt conversion either.

The reason that lowering exposure to recover blown highlights sometimes results in magenta only highlights is technical and has to do with the way digtal camera sensors are made. There are twice as many green pixels as red or blue ones. 60% of the luminosity of the image is in the green channel. Magenta is the absense of green, essentially negative values in the green channel. After there is no more information in the red or blue channels, this is all there is.

wxwax · April 5, 2005

Rutt, a question about step #1. I figured folks did the two exposures/layer mask thing when it wasn't possible to create a single image in which neither end of the historgram was clipped.

rutt · April 5, 2005

wxwax wrote:

Rutt, a question about step #1. I figured folks did the two exposures/layer mask thing when it wasn't possible to create a single image in which neither end of the historgram was clipped.

I haven't seen such an image. Decrease shadow to 0 and the shadows that can be unclipped will be. Decrease exposure and the highlights that can be unclipped will be. The problem is that the areas of interest are all so compressed in the resulting image as to be invisible. I mean that the highlights and shadows, though not clipped are each compressed into relatively small areas at the end of the curve. The shadow/highlight adjustment overlaps different sections of the curve, essentially exposing detail that exits in the representation but which isnt visible.

Andy · April 15, 2005

improving the bass
rutt's been after me to try something, so i did. i followed these instructions:

1. In ACR, make sure neither shadows nor highlights are clipped by using the shadow and exposure sliders, respectively. Do just enough to get the whole histogram in.

2. In shadow/highlight, use shadow amout = 0, and play with the highlight settings. You should be able to recover the detail in the highlight if there is any (and we know there will be because you were able to make two conversions and combine.)

3. Use curves (preferably LAB, L) to restore contrast. Start by just moving the endponts in until you have true shadow and are just starting to lose detail in the highlights. Then screw around. (andy: that's a technical term

)

this is rutt's method, and rutt's sharpening

here's my original edit from two months ago. this edit involved developing two exposures from the raw file, and blending them in post

it's an interesting approach. i'm going to print these shots off 13x19 in the morning to see if there are noticeable differences in the final output. i do like having the extra control over the exposure, and i believe that after working this thru a couple times that i may be an improvement over my existing workflow for many of my images.

thanks for your contribution to dgrin, rutt, and for working with me on this tonight!

rutt · April 15, 2005

And, the ACR+shadow/highlight technique for highlight recovery does seem to work really well. I think the edge of the recovered highlight is smoother and more natural in tonight's version, which was the real point of the excecise.

To my eye, the sharpening from your original is a little more aggresive and looks a little better. Look at the white on black letters behind the base. To print at large size, you'll need to sharpen again with differnet parameters than when we sharpened the resized image. Be a little braver with Radius and Amount. Remember the trick doesn't work if the results aren't visible.
Most people are a little timid about both of these parameters. If you sharpen last before printing and save your work before sharpening, you can always go back and back off the parameters if you have visible halos in the print, so it isn't the end of the world. Try to look a section of the image the very same size you will print it. It's too bad, but nothing beats sharpening for the intended viewing size.

rutt · April 16, 2005

This is an image of Andy's. Originally he processed by converting twice with different ACR parameters and then using a layer mask to blend the two conversions. Here is Andy's version that he posted about a month ago:

Here is what he did in detail:

convert with default parameters
convert again with exposure -.95
layer the two conversions.
use a layer mask to show the part below the sticker half way up the pole with lower exposure and the top 1/2 with higher exposure.

I've given away Andy's magic trick, and now if you look carefully at his image you can see the area in the middle of the image where the blends meet. You'd probably never notice this if you didn't know how it was done, but once you do you can see the transition.

I thought I'd see how I could do with shadow/highlight on this image. I 20/20/10/10/50/0 and followed with LAB curve A+B steepeneing and arrived at this:

You can see that I didn't actually manage to duplicate Andy's result. The top 1/2 is considerably darker than in Andy's version. And this isn't surprising. I've used only transformations that apply to the entire image consistently. On the other hand, the there is no transition between the top and bottom of the image. No matter how hard you look the light is consistent. The sky is also bluer than in Andy's version because of the LAB steepening.

I was able to acheive a result more like Andy's without making two conversions, but by using a layer mask. I duplicated the image above into a second layer and applied an L channel curve to lighten it up a little. Then I used a layer mask to achive an effect similar to Andy's by blending the top half of the lighter layer with the bottom half of the darker layer. Of course, my brush strokes don't exactly match Andy's but you get the idea. Not surprisingly, my transition is about as visible as Andy's.

I guess the moral here is that the shadow/highlight tool can't do everything that a layer mask can. Not surprising. But:

It still isn't necessary to do two conversions. It's sufficient to convert once and then use PS tools after so long as the conversion delivers all the data, and
Local transformations with selections, brushes, layer masks, &etc can produce effects that can't be replicated with global transformations. But unless they almost always leave some evidence if you know what to look for. This doesn't mean that they shouldnt' be used or the the results won't be good. It's just the nature of the beast.

rutt · April 18, 2005

More thoughts on blending Andy's bird on pole shot
Remember this is Andy's shot, not mine.

I thought some more about the right way to get the effect Andy wanted for the seagull on the pole shot. There are two conflicting goals for this image. Andy wants to lighten the image on the top so that the bird is more visible and has visible highlights so it looks as if it is catching the last (or first?) of the suns rays. And he wants the dark shoreline and as much detail around the sun and in the clouds in the bottom of the shot as possible.

At the end of my last post, I concluded that I didn't know how to blend a lighter top of the shot with a darker bottom of the shot without using some local moves and thus some evidence for the careful eye. I came up with one way of doing this and I'm going to explain it in some detail.

The idea is to use the gradient tool to paint in a layer mask and thus make a very smooth blend. In fact, I drew a gradient from the top of the shot all the way to the bottom, so the blend can actually be considered a global move instead of a local one and I think the appearance is very smooth.

Here is my final result:

And here are the steps I followed:

Convert from raw with ACR. As described throughout this thread, I adjusted both exposures and shadows to make sure that nothing was clipped. This isn't quite true; in this shot, the sun itself is blown and needs to be blown and there is no recovering it, so there will be some highlight clipping. But we want to recover as much as possible. Here is the ACR dialog for this:

and here is the image immediately after conversion:
I made a pair of duplicate layers. The top layer is for the darker version and is optimized for the bottom of the image. The middle layer is for the lighter version and optimized for the top of the shot. Oh, by the way, at this point I duplicated Andy's crop. He cut off the very top of the image, just enough to remove the airplane in the upper left and give the image slightly more balanced proportions.
I applied an L curve to the top layer. My goal here was to reveal even more detail in the clouds on the bottom of the shot and increase the contrast of the coastline. Here is this curve:

and here is the resulting dark layer:
I applied a different L curve to the middle layer. My goas here was to lighten it and expand the midtones to highlights to make the sky much lighter and give the bird the illusion of being bathed in "sweet light". Here is the curve:

and here is the resulting light layer:
Neither of these is what we want. The light layer blows a large area around the sun. The dark layer is much too dark around the top. So we need a good way to blend them. Like Andy, I added a layer mask. Unlke Andy, instead of using a brush, I decided to use the gradient tool in a layer mask on the topmost (dark) layer in order to shade it smoothly into the middle (light) layer. I used the linear gradiant and it took a little fussing to get the gradient options right so the blending would happen is just the part of the image I wanted. Here is how I setup the gradient dialog box:

I then drew a vertical line with the gradient tool from the middle of the top to the middle of the bottom of the image. Then the topmost (dark) layer looked like this:
With both middle and top layers visible, I got the result displayed on top of this post.

At this point, there are a lot of things we can do with the image. For example we can increase saturation of to get a bluer sky using LAB A+B steepening. We could sharpen (though this particular image doesn't have anything that really should look sharp.) Both of those things should be done after flattening. Before flattening, we can play the separate layers in various ways as well, fine tuning the L curves, perhaps. For example, I applied LAB A+B steepening by moving the endpoint of both A and B curves inward by 10:

Lessons:

I still didn't have to use two different conversions. By making sure that the conversion I used clipped no details I wanted in the final version and by converting to 16 bits, I was able to make light and dark exposures with PS L curves. Curves are much more powerful than the ACR sliders and since there was no loss of data during the conversion, I was able to acheive light and dark exposures of equal quality and with more control than possible with ACR.
By using the gradient tool to paint in a layer mask, I was able to acheive a global effect that blends very gradually between the layers. In my final version, there is no single area of the image where the blend takes place and so no tell take clue of what happened for even the informed viewer.

rutt · April 18, 2005

Looking at my final versions vs Andy's, he was braver about the lightness of the top of the image than I was. This doesn't change the point of my post or the lessons learned there. I just needed to be braver the the lightening curve.

Perhaps with this much lightening of the top of the image, both the gradient and the curve for the bottom half also need readjustment. I don't think the sky on the top of the image should actually be lighter than the blue sky just above the sunset, do you? I think this was a problem in Andy's original edit as well.

Mike Lane · April 19, 2005

rutt wrote:

It seems that a lot of us know how to use image->adjustments->shadow/highlights to recover shadow detail, but it is equally useful for recovering highlight detail. I've noticed quite a few messages over the last month or so about making two raw conversions and then recovering highlights by using a layer mask and painting out the blown details to allow the darker image to show through.

There is a global (no PS brushes or selections) alternative to this that can result in very natural looking results and which requires no skill or time with the wacom tablet. Essentially:

There is a way to combine two images (an ACR image extracted once for highlights and once for shadows) to get better highlights and shadows. I found this on the Adobe Tips and Tutorials site so please forgive me if this isn't new to anyone. Also sorry that this goes so in depth on basic things.

http://studio.adobe.com/tips/tip.jsp?p=1&id=408&xml=phs8contrast

Total Contrast Control Using Two Images

Skip to subnavigation

George DeWolfe. Adapted from The Digital Fine Print Workshop by George DeWolfe. www.georgedewolfe.com. George DeWolfe is Senior Editor for Camera Arts and View Camera magazines and teaches digital photography workshops worldwide. He studied with Ansel Adams and Minor White and holds an MFA in Photography from the Rochester Institute of Technology. His prints are in major national and personal collections.

Contrast is the supreme technical problem in photography. In traditional print photography the major limiting component is the range of the paper, or 5 f/stops of detail from highlight to shadow. It was for this reason that the Zone System was invented—to manipulate the exposure and development of the negative to fit the print. In the darkroom there are also different contrast grades of paper or different variable contrast filters to adjust contrast. In Adobe® Photoshop® CS, I use the following technique to correct and adjust the overall contrast of digital images. The technique uses 2 digital files from a digital camera mounted on a tripod, one of which has been taken for the highlights and the other that has been taken for the shadows. The 2 images are combined together in Layers using a Layer Mask and Curves into one seamless image containing the highlights of one, the shadows of the other, and a good blend of midtone values.

1. Combine the 2 images and crop.

Open both underexposed and overexposed images into Photoshop. Click and drag (using the Move Tool while holding down the Shift Key) the underexposed image on top of the overexposed image. This puts the underexposure on a layer on top of the overexposure (which becomes the Background). In the Layers palette, select Layer 1 and set the Blending Mode to Difference and align the 2 images with the Move Tool, if necessary.

Crop the images here. (Cropping the image here at the first step keeps the images aligned if you have moved Layer 1 even the slightest amount). Save the image (File > Save As) as Combo and close the 2 original images without saving them. Return to Normal Blending Mode.

2. Make a Layer Mask and paste the background layer into it.

On Combo make a Layer Mask (Layer > Add Layer Mask > Reveal All) on Layer 1. Select the Background Layer (Select > All) and Copy (Edit > Copy). Hold down the Option/Alt key and click on the Layer Mask (the document window should turn white). Paste the Background Layer (Edit > Paste) into the Layer Mask (it will be pasted as a Black & White mask).

3. Make a new window.

Go to Window > Arrange > New Window for Combo.psd to see the effects of the mask. You will not be able to see the actual picture because you will be working on the mask. This is the reason for making the New Window.

4. Blur the mask.

Click on the Layer Mask and Gaussian Blur (Filter > Blur > Gaussian Blur) the mask to a 5–50 pixel Radius. You want to blur enough to get rid of all texture and detail.

5. Adjust the mask contrast with curves.

Select Image > Adjustments > Curves and adjust contrast and brightness of the mask until the image looks correct in the New Window. You do not want to use a Curves Adjustment Layer here because you are still working on the Layer Mask. Close the New Window after you’ve finished adjusting with Curves.

6. Flatten the image.

Deselect and Flatten the image (Layer > Flatten Image).

7. Conclusion.

The great thing about Photoshop is that we have much more control in adjusting the overall contrast than we do in traditional photography because we can use both Black & White and Color images. This technique frees your photography and enables you to capture any extreme lighting situation with ease.

Andy · April 19, 2005

rutt wrote:

Looking at my final versions vs Andy's, he was braver about the lightness of the top of the image than I was. This doesn't change the point of my post or the lessons learned there. I just needed to be braver the the lightening curve.

Perhaps with this much lightening of the top of the image, both the gradient and the curve for the bottom half also need readjustment. I don't think the sky on the top of the image should actually be lighter than the blue sky just above the sunset, do you? I think this was a problem in Andy's original edit as well.

rutt's

my orig

i dunno, rutt - you got more green in the sun area, and lost some of the original reds there... also look at the birds feet/legs and beak ... you've lost some of the saturation there, too.

i dig your creativity with the gradient, but i think my orig edit (3 mins total time) was better .... can you explain to me how gradient mask is any different that two different exposures from the raw file and blending? i'm not following that part of things..

great stuff!

rutt · April 19, 2005

Andy, my goals here are:

To avoid making two ACR conversions,
To blend as naturally as possible, and
To avoid any step that requires manual dexterity.

1. is sort of a theoretical thing. I know that 16 bit conversions without highlight or shadow clipping contain all the data. I want to figure out how to expose it.

2. is the real payoff here. Once I learned the technique you used to blend, I could see it in you image, even at smugmug L size. The two halves of the image don't fit organically together. I'm trying to figure out how to make this work better. To be totally honest my two halves don't blend together perfectly organically, but in principal they could. The gradiant would just have to be completely even from the top to bottom of the image. I haven't done this because I was too concerned with matching your colors. (More about this below.)

Here are the two images. I've drawn some lines to delineate the areas where the blend is visible to me:

Andy's:

Rutt's:

3. is has to do with my own strengths and weaknesses. I am good at figuring out how to do stuff and not so good with my hands. For me it's worth a lot of thinking to learn a new technique that I can apply autmatically in the future. If I nail this blending technique, it could even be recorded as an action and used whenever there was an image with very different dynamic range on the top and bottom.

OK, a big hard topic that I'm not really good at is color matching. I've benn trying (and failing) to match Andy's colors as closely as possible while also exploring how to make the most organic blend possible.

Keeping my own strengths and weaknesses, I did this one more time using the same basic procedure, but this time my primary goal as to show how indetectable I could make the blend and secondarily how closely I could match the darkness at the bottom of Andy's image and the lightness at the top. Here it is:

IMO, this image has a very indetectable blend. I'm not the artist; Andy is. Probably I haven't expressed exactly what he was after. But as a technical exploration, I think this shows the viability of the technique.

XO-Studios · April 25, 2005

rutt wrote:

Andy, my goals here are:
To avoid making two ACR conversions,

To blend as naturally as possible, and

To avoid any step that requires manual dexterity.

<SNIP>
IMO, this image has a very indetectable blend. I'm not the artist; Andy is. Probably I haven't expressed exactly what he was after. But as a technical exploration, I think this shows the viability of the technique.

For another interesting way, and to make multiple adjustments in a picture with various exposure challenges on top of indetectable blends (use proper amount of feather) here is a post from another thread.

http://www.dgrin.com/showpost.php?p=95201&postcount=3

Enjoy,

XO,

Recovering highlights with shadow/highlights

Comments