Blending Multiple Exposures for Dynamic Range and Depth of Field

My Singh-Ray GND's are scratched beyond belief and instead of dropping another $300 to replace all of them I'm thinking I might just start hand-blending all images. I understand the basic steps, i.e. one highlight and one shadow image, stack them, use a brush to paint appropriate areas, etc. Simple enough in concept. I've used the technique a couple of times and have found that my bird brain is getting caught on a couple of items.

Blending For Dynamic Range:

I'm getting a little lost on brush and feathering choice. Do you first use broad strokes and a heavily feathered brush followed by a smaller brush with low to no feathering to clean up "spills"? That's what I've been doing and it's rather time intensive. Is there any easier way?

Also, how do you blend more than two expsosures? For simplicity's sake, let's say you have 3 exposures. Do you first blend two of them, then blend the third with the result of the first blend? Is there even any reason for using more than 2 exposures (one shadow, one highlight)? I understand why several exposures are better for automated HDR photography but I'm not sure I understand why the same would be true for a manual blend.

Blending For Dynamic Range AND Depth of Field:

Now let's assume you want to blend images to control dynamic range and achieve loads of DOF. In what order should this be done? Blend images for DOF first, then blend images for dynamic range? Does it matter? If so, why does it matter? I'm not being difficult. I just really want to get a handle on all this. I'm starting to get a lot of questions about both at my workshops and I need to be able to speak intelligently on the topic(s).

I apologize for the lengthy post. Looking forward to hearing your thoughts!

I'm getting a little lost on brush and feathering choice. Do you first use broad strokes and a heavily feathered brush followed by a smaller brush with low to no feathering to clean up "spills"? That's what I've been doing and it's rather time intensive. Is there any easier way?

Before I begin 'painting' on the layer mask, I select the object(s) I want to blend in or select the opposite area and Select->Inverse (whichever makes for the easiest selection). Then I feather the selection by 1 or 2 pixels, and sometimes contract (or expand) the selection by 1 pixel in order to avoid ragged edges. Choosing a selection first allows one to 'paint' with abandon and eliminates the need to be ultra precise when 'painting' along those areas.
By the way, this is how I target areas to clone or heal along edges for the same reasons.
Make sure to save the selection after you finish so you don't have to recreate what can be a tedious chore at times.
I've never needed to blend more than 2 exposures for expanded dynamic range (knock on wood )

I look forward to more discussion on this topic and the forthcoming tips, so am real glad you raised these questions.

I never thought of making a selection but that seems like an excellent technique for most situations. I tried blending a waterfall image from the Tetons not long ago but gave up because there were so many unsmooth edges, i.e. dark tree limbs poking into the bright waterfall area. In that situation I don't think making a selection would work as well. Thanks for the tip, John. Can't wait to give it a shot!

My pleasure, Bret.
Without seeing the waterfall scenario, it sounds like the much maligned Magic Wand tool would have been perfect for selecting the bright waterfall against the darker tree limbs.

I failed to mention that Tony Kuyper has a superb tutorial on blending images using his self feathering Luminosity Mask curve adjustment layer masks titled 'Hand Blending HDR Images Using Luminosity Masks' with a three step summary at the end.

Bret, three quick notes:

1. Exposure blending for some bracketed sets (and for some differently-converted-from-the-same RAW file) can be done in 30 seconds by this method posted on Luminous Landscape. Really. Though I can't leave good enough alone and usually fuss with the two layers a while longer.

2. HDR Shop (version 1) is freeware, stable, and 800 kb; it will blend 2, 3, 4, or more images at once. Output as a normal (not specifically HDR) file type is usually worth the time it takes, and the results have a little more range and a little more contrast than the 1st method.

3. Am I backwards or weird in finding the usual way of treating layer masks (make a Hide All mask and erase with a white brush) much more difficiult than making the mask a Show All mask and painting black on what I want to hide? I can't see through the ruby overlay well enough to be confident about tree limbs, etc., and spend too much time clicking the overlay on and off just to see what I'm doing. Give the "show all" and paint-it-black a try.

Bret,
Are you using Lightroom? If so, give LR Enfuse a try. It's not 100% perfect, but it's saved me lots of time with blending in PS. Also, Photomatix has a blending function called Fuse Exposures. Sometimes these tools can save you a bunch of time right up front, so you won't have to spend a lot of manual time in PS. Here's the link for LR Enfuse -
http://www.photographers-toolbox.com/products/lrenfuse.php
For PS work, I almost always start with just two images, add a layer mask to the top one, then use the Grad filter tool to start, then finish up with a brush if necessary. If I need something from a 3rd image, I put that one at the top of the layer stack, add a mask, fill it with black, then paint in whatever I need from the combined images below.
As far as when to do DOF, I find that doing coarse adjustments first (DOF) then the finer adjustments (blending) works better for me. In this case, (and I use either CS4 or CombineZP for the DOF) I find it easier to work with images of the same exposure when I do DOF work, then worry about exposure blending later. And, although you didn't ask about HDR panos, I found that Photomatix likes smaller file sizes, so you're better off creating individual HDR files first, then creating the pano second. I hope this helps.
Paul

A Comprehensive Article for increasing dynamic range.

From my honest point of view, it's a time consuming thing. More than 2 images even more time. Some are easy but some are harder. I sometimes work with lightest to darkest or vice versa. Sometimes I use Tony Kuyper's masks. Once you get them blended you'll need to adjust contrast in specific areas most of the time. Double raw processing needs all of this as well. As for grads if stuff is moving they are great to have. I use the hard edge at the beach a lot. The soft edge for many other things while it can darken things I don't want it to it still can retain a little bit of contrast on the horizon which can make blending easier some times.

As for the brushes, all different kinds and all different kinds of selections too.

Generally, I tend work with the exposure that has the most data I want to keep which is most of the time the dark side of the histogram.

If you're mainly looking to replace GNDs but accomplish a little that they can't (avoiding objects sticking into the grad line, etc.) this is simple and pretty quick. Here's a good tutorial on that Exposure Blending Basics, which is almost exactly how I do it, with a couple of exceptions.

-- I tend to use the grad tool rather than a brush to set the initial light and dark parts of the image - replicating pretty much exactly what you'd get using a GND in the field.
-- Or sometimes I do a selection - say a trace around mountains and foliage - then run a gaussian blur on the layer mask to make essentially a custom GND

Then it's a matter of tweaking to taste.

Hi Bret,

95% of what I do is with a brush on the layer mask. The absolute key to doing this quickly is to use shortcut keys for brush size and brush feathering. To change brush size, use the [ and ] (bracket) keys. To change brush feathering, hold Shift and press the bracket keys. To replicate an ND grad effect, it takes only a few seconds.

If you find your edges are too visible, apply some Gaussian blur to the mask: hold Option (mac) and click the layer mask (this reveals your mask only), then go to Filter > gaussian blur and apply the amount (often between 50 and 150)

John: I forgot about Tony's technique. I REALLY need to dig into his tutorials and actions. Thanks for the reminder!

Paul: I am using LR. I used Photomatix for a little while and rarely ever received results I found to be acceptable. That, and it seemed like the software always softened details that were razor sharp in all of the original source files. It's been about a year and I see that there is a newer version available. I may download it and see if I get better results. The Fuse Exposures feature is new, too.

David: Thanks for the tutorial link. Very helpful!

Doug: I do need to get used to using PS shortcuts more often. It seems that blending is a perfect reason to use them more as it is quite time consuming to change brush sizes/opacity by point-n-clicking. I appreciate the tips!

Thanks for all the good information thus far. Keep those tips, tricks and techniques coming!

I use John's method. In theory it would work well with Tony's masks, but I find that the luminosity masks are too feathered for blending, but sometimes offer a good starting point (they're super for curves adjustments for fine tuning). I use the magic wand if I want crisp edges like along a ridge, or the select color range tool for other situations. That tool is amazing!

Floris: I was hoping you'd chime in as you seem to do quite a bit of blending, and you do it very well. Thanks for sharing your thoughts.

I use a hybrid process: Using photomatix (with some really forgiving setting so that it does not look HDRish) I create a blend from 2 or 3 or 4 (whatever) no of frames that I have. Then I use that as the base layer and drop all the frames on top as layers. Then, using selection tools and brush I paint back the areas that I like in different frames - having the opacity less than 100% (50-70%) range is the key for good transition.
New version of Photomatix is really good - it has a separate "Exposure blending" tool which produces much better result than the "HDR" tool for my purpose

hope this helps

AND yes: I am not buying those expensive GNDs anymore - like you. They are a pain to carry too - and so easy to scratch. I am moving on (though I still like certain advantages of GNDs - trade off in life as always )

Arnab: I made my decision to stop using GND's when I dropped an almost new Singh-Ray onto some pea size gravel from about 6'. Yeah, it scratched the crap out of it. It's about time I get with the current technology!

Color Range selections are your friend, for the manual process. I use them quite a bit for tough edge situations. I've been using a very basic technique for many years. It's probably one I won't give up on easily either. Just works, and I enjoy it. I start with large feathered brushes (image dependent, but generally 1000-2500 pixels) and jump in at 50% opacity or less. I build up the mask slowly (as in a 5 minutes, instead of one minute with a gradient). And then my brush size keeps getting smaller as I refine. I will often use Color Range selections for fine tuning and/or high contrast edges. I never work with more than two exposures.

Like Arnab, I also occasionally make use of a Photomatix tone mapped layer as a final step, with low opacity blending. HDR software is finicky, but subtly blending in a tone mapped layer can really help with the dynamic range - yet still avoid the HRD look. Anyway, just some random thoughts off the top of my head. Color Range selections and quick keys (like Doug said!).

Oh, and I am still not adventurous enough to try out exposure blending AND focus blending in one image. I'll do the focus blending if the exposure isn't challenging. Otherwise, I stick to my DOF chart.

I have actually tried to do focus and exposure blend at the same time - but have not mastered the workflow yet. As obvious, your focus and exposure should match (e.g. same frame should be exposed and focused for FG or whatever) - it's a little difficult to achieve when light changing fast, you want to find the right composition etc etc. I set my camera to take 3 exposures in the order of +1 0 -1 (or +2 0 -2) and then go from FG to MG to BG (as generally FG is dark and BG is illuminated in most landscapes - otherwise you have to change based on the scene).

So, you have to get the process right during the capture - the blending process should be similar in both cases.

To make it more complicated, add stiching to it - then you are on to something

When you guys use Color Range selections how much feathering are you using? I realize it's going to vary but if there's a good starting point, I'd sure like to know it. Do you have a couple images you'd be willing to post as examples w/ a brief explanation of how you processed the image?

I rarely use global blur/feathering, as I don't like to mess around with the edges. Color Range selections are very accurate, and I rarely see artifacts, halos, etc. when using this tool. The most I will do is hand paint (with the blur tool) on the mask itself, to selectively soften mask edges ... usually within a highly masked layer devoted to the dark shadow edges. I also really like the Fade menu command - which provides that much more control over the manual painting process. For example, select your brush and your opacity, and then paint away. Then use the Fade command (0-100%) to fine tune the effect of your painting.

Hey Bret, next time you have a tricky shot, get in touch. I'd be happy to give it a shot with my approach, and walk you through my solution.

One thing I've been doing more of lately is creating masks from channels. It's fast, easy, and makes some pretty sophisticated masks.