This content has been marked as final. Show 20 replies
For truly high dynamic range images you need to use local tone mapping operators which means some form of HDR merging. A simple tone curve is never going to do the trick, because if you expand in one zone you will compress in another. You need local contrast.
This is the reason that landscape photographers use split neutral density
filters. It's easer to fix many of these problems before they get onto the
chip. Alternatively you could take two shots one for the highlights and one
for the shadows and blend them in Photoshop. These are several of the
methods that I've heard of to deal with this problem.
The process is called growing as a photographer.
I've used graduated filters in the past, but I don't like the results with very bumpy skylines.
I've also used exposure blending of two prints of the same raw file, but still haven't worked out a way of doing this in under an hour, especially where the sky/land boundaries are diffuse.
Hopefully I am growing as a photographer, but it's always good to get advice from more experienced and knowledgeable people - especially those who use ACR - so that I don't waste the next few months using the wrong post-processing techniques.
>Can anyone suggest a better way of using ACR with high dynamic range captures?
The big challenge at this point is trying to get a single tone map from Camera Raw since there is no capability to do local adjustments. However, take a look at the example in the book that shows the Scottish castle and uses Smart Objects to blend 3 different CR settings of the same image.
While true HDR photography can indeed be used, there are issues of control over the toning and limitations of the tool set to work with. I much prefer working with a multiple exposure bracket that is smaller (say three stops instead of 9-10) and use layers to blend the results. Using blending options and layer masks you can put the tone you want where you want it.
The thing that bothers me about HDR is that it's not particularly necessary in the photos I'm converting, as I'm not dealing with huge extremes. So I feel that it's an unneccesary amount of post processing. I am basically lazy, and I always look for the simplest solution.
My camera supposedly records 14bit raw and has good noise performance; a lot of my conversions involve Shadow fill and Recovery settings of 30, so the midtones are bound to get squeezed, but I'm not getting too much shadow noise as a result, and don't feel the need for extra captures. So my current workflow involved tonal compression with ACR and then expansion with a Local Contrast filter.
I agree that some form of HDR is required to avoid the initial compression, but I think the nail was hit on the head about the lack of local adjustment in ACR. I suppose if I had that, things might be different. Maybe I should (spit) take another look at Nikon Capture NX - no I can't bring myself to do it.
One way to do it that works on some pictures is to start by turning down contrast. Then you use the point curve as follows:
Ctrl-click in areas where you want contrast. Put another point close by and move the 2 points ever so slightly. This way you can sometimes get enough contrast in the places that need it while keeping total range.
Sometimes this can be a really good for getting a pleasantly low contrasted picture that still has sting. Very nice in these times of overly contrasted pictures.
This is actually the only thing I miss in Lightroom that makes me go back to ACR.
For a good philosophy background on making good pictures of high range scenes, look at:
>The thing that bothers me about HDR is that it's not particularly necessary in the photos I'm converting, as I'm not dealing with huge extremes. So I feel that it's an unneccesary amount of post processing. I am basically lazy, and I always look for the simplest solution.
Then why don't you use the Smart Objects blending method suggested by Jeff Schewe? In the case of the Scottish castle, it was easy to make a mask to separate the highlight from the shadow areas, since the highlights were in the upper half of the image and the shadows were in the lower half. If the highlights and shadows are intermixed and the manual creation of a mask would be difficult, you could try loading luminosity as a mask as discussed by Bruce Fraser et al in the Real World PSCS3 book (page 462). You can also use Smart Objects here.
Another approach would be to convert a 16 bit image to 32 bit and use the HDR tone mapping controls.
>Then why don't you use the Smart Objects blending method suggested by Jeff Schewe?
I've done exposure blending of multiple prints before, with partial success. My main problem was creating a decent mask. I'll going to give Jeff's example a try; I found the page this morning, and will give it a spin when I've got a spare hour or two.
>Another approach would be to convert a 16 bit image to 32 bit and use the HDR tone mapping controls.
I've tried using Photomatix Pro's Tonemapping feature on a 16bit TIFF generated by ACR. I didn't convert to 32bit though. I've not been happy with Photomatix's results so far - I often get terrible grey patches. I thought I was doing something wrong, but an HDR expert checked my settings and said they were reasonable. I assume my subjects are just unsuitable for tonemapping.
>I've done exposure blending of multiple prints before, with partial success. My main problem was creating a decent mask.
Have you tried luminosity masking?
Okay, I've spent the last couple of hours experimenting with smart object blending, luminosity masking and tone mapping. I worked on a couple of images, and then compared the results with the same image processed with my current workflow.
The new techniques I tried took much longer than I usually spend on an image, and produced inferior results. I preferred the images created by compressing the raw into Photoshop and then correcting with small amounts of local contrast. The blended photographs appeared dull and artificial in comparison.
The images I chose had 2 stops difference between light and dark portions. One had water, one had diffuse trees/grass. They looked okay until I compared them with the old system, which seemed superior in most respects - detail, balance, finish.
I was expecting the blended images to look more natural, but it was the opposite - presumably because of the big difference in exposure between different parts. I tried smoothing the transitions, but this made things worse, producing either halos or reducing local contrast.
I was very impressed with the Luminosity Masking technique, which I'm sure I'll use more in future. I'll have to find a way of refining the selection (I used Contrast/Brightness Adjustment as well as Gaussian Blur).
Anyway, thanks for everyone's help and suggestions. I think I'll stick with my current system for the time being. If it ain't broke, don't fix it.
Sounds like you are ripe for my earlier message (#6) ;-).
At least the link at the bottom which might help, whatever technique you choose.
Hi John. I did read your post and followed the link.
The Curves method was the one I used before discovering Local Contrast filters. Jeff/Bruce's book showed me how to pinpoint tones in the curve using the Ctrl key, and I made slope adjustments at key points in the curve. Again, this is a general adjustment across the whole image, and doesn't take into account the 'local' nature of image perception.
I could experiment with Curves Adjustment Layers with Luminosity Masking in Photoshop, but I can't bring myself to do this when a filter will do this for me in a few seconds. ;-)
Wish you could post a sample of one of your images to show what you are talking about.
Not sure if what you're after is definition in detail like in clumps of foliage or trying to prevent blown highlights and plugged shadow detail while retaining eye pleasing contrast throughout.
I use a HDR test image I shot of a brightly lit midday scene containing big fluffy clouds, trees at various distances, a limestone wall receding into the distance, carpet grass and telephone poles and street signs. I get so much detail without anything blown out or plugged up just using curves in ACR after first exposing to the right during capture.
I study the histogram's peaks and valleys and form flat areas where the peaks are on a subtle sloping S-curve and carefully tweak different spots on the curve to add definition in the shadows of the tree foliage and carpet grass. After doing this I usually don't even have to sharpen in PS.
If you have a big hill or triangle shaped peak in the middle and another close to black point in your histogram you might form a gentle sloping M-shape curve creating a three point wide pulldown in the middle that makes the entire curve look like a gull wing. Flatten portions of the curve in an angled slope to add definition to the 1/4 and 3/4 tone regions.
Of course I have no idea what look you're going for. All I go for is what I saw through the lens and I can get really close just using curves. I may spend a lot of time on forming this curve but I end up saving it and applying it to similarly lit scenes with very good results.
I have a selection of curves I use for high contrast scenes, except mine are upside-down Ms. They are more for correction to low contrast highlights and shadows (which have peaks at the left and right) than initial shaping, which I do through the Basic panel.
Here's some images to illustrate what I've been talking about.
Here's two photos taken with my old camera, using Basic and Curves:
Here's two fixed with Exposure Blending:
Here's two done with HDR (the rest got thrown away):
And here's two done with Local Contrast enhancement:
Just for clarification - when you say "Local Contrast enhancement", are you referring to the technique with an unsharp mask large radius, or do you mean something else?
Your postings don't explain what you're after.
If you could just post a before and after of the SAME image using one technique over another, it would make it more clear. Having various scenes and using different techniques makes it hard to determine the effects of anything due to the variances in exposure, lighting and distance from subject.
Also the size of the images are too small to tell what you're illustrating.
Frankly your first two images look the most natural of the bunch but I can't tell if it's due to the overcast outdoor lighting or any contrast adjusts you've applied. But overall they all look good and natural looking, some more than others.
And just to add.
Our eyes expect subjects that are closer to us to have more contrast and sharpness compared to objects that are in the distance.
One thing that kills a landscape scene using what some consider HDR techniques is that they apply global and local contrast and sharpness equally to all the objects in the scene both foreground and background. This is what gives that fake look as if everything is all on the same focal plane of sharpness. Our eyes just don't perceive distances that way.
Your "Rowers On Lake Windermere" is of this type but not as pronounced.
Your HDR examples would be good in my eyes if the skies were a bit lighter. I feel most people overdo this. Which also applies to graduated filters ;-)
But of course this is a matter of taste.
>Just for clarification - when you say "Local Contrast enhancement", are you referring to the technique with an unsharp mask large radius, or do you mean something else?
It's along those lines. I've used that method, and it produces similar but cruder results than the Nikon Photoshop plug-in called Tonal Contrast, which actually works on three separate tonal bands individually, presumably with physical boundaries decided by average luminosity.
The Venice photo has a very big dynamic range (bright high direct sunlight and dark shadows. But I managed to bring detail back to the whole image with this filter, without it looking artificial.
In answer to another question: I'm trying to achieve a good level of contrast and detail throughout an image, because I believe that we perceive detail subjectively on close examination, and high dynamic range images often suffer from low contrast. It's all about artificially approximating perception, which no doubt varies between people.
I like the HDR/tonemapped photos in a fake sort of way, but the techniques worked on very few of my photos and I generally produced awful results. I do like some HDR, when it's done well, but I can't seem to master it myself. I agree about the relative brightness problems, which you can also get with exposure blending.
I partly disagree about the relative contrast concept. Okay, in reality, yes, we expect less contrast from far objects, but images are 2-dimensional low-contrast representations of reality, so a bit of creative licence is necessary to recreate the scene as perceived by the photographer. I maintain that contrast is increased by the brain, when examining individual objects in real life, so this can be recreated in post-processing to compensate for the lack of overall contrast.
I'll attempt to do a before/after series, using several techniques on the same image, when I've got more time.
Rephrasing the original question, what I was really wanting to know was if my current workflow was inefficient, in terms of data integrity and effort to achieve the results outlined in my last couple of posts? At the moment, I'm happy with the results, but I'm always on the lookout for 'a better way'.