A more difficult image - wrong time of day
jjbong
Registered Users Posts: 244 Major grins
Here is a picture of the Coloseum in Rome:
Potentially dramatic against the sky, but the wrong time of day. You hardly see the Coloseum. Maybe there's a cast there, too. The best thing would have been to take the shot when there was a southwest or west sun. But for us non-professionals, we're often in the right place at the wrong time, and we don't have the option to wait for the right conditions. I think even professionals have this problem in some cases.
So let's see how we approach this in PS and in other tool sets.
Potentially dramatic against the sky, but the wrong time of day. You hardly see the Coloseum. Maybe there's a cast there, too. The best thing would have been to take the shot when there was a southwest or west sun. But for us non-professionals, we're often in the right place at the wrong time, and we don't have the option to wait for the right conditions. I think even professionals have this problem in some cases.
So let's see how we approach this in PS and in other tool sets.
John Bongiovanni
0
Comments
When I am faced with this I would meter for the coliseum and let the sky blow out, or take a number of bracketed shots for an HDR.
That said here is a quick try on the small image.
Sam
My go to move for exposure issues is B&W. I converted to B&W then duplicated the layer. The duped layer I inverted and changed to soft light blending mode. I did that again and on the second duped layer I played with the opacity slider a bit until I was happy.
And there is my 2 minute fix...
But one of the major rules of Graduate School, is that one needs to quite specifically outline the steps you take, so that someone else could duplicate them. General descriptions are not adequate. I feel this rule is very important here in Grad School, so that everyone can try them steps out themselves, on their own images.
I encourage Sam and Toshido to incorporate that information in their posts, so that they can remain here on Grad School. Other wise I will have to take them down, and I really, really do not want to do that, but want to see the specific steps they took, so that I and others can learn from their editing. My comments are meant as an accolade, not a criticism. Toshido did include an general outline of his steps, but I would like to see more details if possible. For myself, I believe the conversion to B&W is done late in my image edting, not early. I am interested to hear the reasoning for converting to B&W early in the process.
John, is this file available as a RAW image? Or a ZIP file of a RAW image? If so, I can make it available on my iDrive so everyone can get a copy if you can snail mali it to me also.
I think this image, if made available as a RAW file, will be an excellent example of the power of Smart Objects. As a jpg, it will not offer the editing headroom it really needs and deserves.
I am sure, John, you are not the only tourist who has captured an image at the wrong time of day, but at the only time available on their itinerary. I have lots of them.
Moderator of the Technique Forum and Finishing School on Dgrin
Moderator of the Technique Forum and Finishing School on Dgrin
n Camera Raw, the only thing I did was set the white balance from a WhiBal card shot, and
move the Black slider to 0 to get rid of the plugged points.
Recall the PPW workflow: Get the colors right, get the best B&W as a luminosity layer, and
make the colors more vivid.
Get the colors right:
Looking at the numbers for the high clouds and the sky, this is way too yellow (i.e. not enough
blue). The white coulds are showing LAB B values as high as 8, when you expect both A and B to
be near 0. The sky is too Cyan (too much green relative to blue, or alternatively not enough
blue). A curve in color mode fixes this. I curved both the Red and Blue channels, with
most of the work in the Blue. There is a little pink in some of the coulds in
the lower left of the image, but in an early morning shot this is probably OK.
Get the Best B&W as a luminosity layer:
Before anything else, I needed to lighten the Coloseum. I used Dan's false profile, multiply
through mask technique. I assigned (not converted to) a profile that was based on Adobe RGB
(which is my working profile) but with a gamma of 1.4. This lightened everything. I then
went into LAB, created two copies of the background layer in multiply mode, merged them, and
added a layer mask. The layer mask limits where the multiply happens. In this case, we only
want the multiply happening in the sky, so the layer mask came from the B channel of the
background layer, inverted and level adjusted, surface blurred and blurred slightly.
Channel Blending: Blue into Red, darken mode. Green into Red, darken mode, 70%.
Curve the Green and Red channels to lighten and bring more contrast into the Coliseum, but at
the same time retain the contrast in the sky. Tried Shadow/Highlight, but wasn't happy with
the results - it got rid of the contrast in the Coliseum. At this point, I was reasonably
happy with the Coliseum, but not the sky. So I did a multiply in RGB using the same mask I
had earlier used in LAB, with layer opacity on the multiply layer at 62%.
Make the colors more vivid:
All in LAB. Use an extreme AB curve at low opacity (Dan calls this the Man from Mars move),
the opacity being 22% here. This brings some color variation to the relatively monochrome
sections of the Coliseum. Then the standard AB curve with opacity 45% to boost the colors overall.
Set the highlight in the high cloud on an L curve, and put a tiny bit more contrast into the
Coloseum.
Sharpened, and done.
I'd like to be able to say that all of that took me just 5 minutes. It took a bit longer than
that, maybe 15 or 20. Getting the color right was easy. Once I lightened the Coloseum, it
was all pretty quick, too. But I had to do a fair amount of experimentation to find the right
way to lighten it. I experimented with various parameters (e.g., gamma of the false profile,
whether to multiply in RGB or LAB, how many layers to multiply, etc.). Like everything else,
it should get quicker with more practice.
I will say that the result of this, which I did from scratch for this forum to make sure I
had all the steps, was significantly better than my original effort.
Two things I didn't do. One was remove the sensor dust spots in the sky (I did this when I
developed the image the first time, for myself). The other was deal with noise, which I also
had not done before. If you look closely, you'll see noise in the dark areas of the Coloseum.
For the size image I'm working with, I don't think this is important. For a larger print, it
might well be.
I apologize for the small image. I took the raw into PS, and saved a maximum size JPEG. I then uploaded it to SmugMug and referenced the XL image. But for some reason, it references (and downloads) a small image. I'll work to try to get that fixed.
You can download the full-sized jpeg here. Right-click and Save Image As (Windows, don't know Mac). Or you can download the Raw file from Pathfinder's site, described in an earlier post.
http://jjbong.smugmug.com/Photo-Work/Dgrin-Grad-School/MG3190/1118833842_2uYCa-O-1.jpg
Apologies again. I didn't understand how SmugMug worked.
Getting the right color balance can be hard in jpgs as so much data has been discarded from the original Raw file conversion down to the 8 bit jpg. But I though "lets give it a go".
I looked around the file, for a neutral color, and tried the clouds, but did not like the results. I finally tried on the white stone cornice in the very center of the image, on the right side of the corner just to the left of the upper arch. Bingo, now the colors begin to fall into place.
The clouds do remain a bit pink, but this is global editor, not a local editor, and I will choose to think that the clouds are reflecting the red brick walls and floors of Rome.
I set the Exposure slider at +0.6, Recovery at 0, Fill Light at 68, and Blacks at 6
I left Brightness unchanged at 0, and also Contrast at 0. I raised Clarity to +21, Vibrance to +53, and Saturation to +38. These are high numbers but I wanted to try to at least match the color densities in John's image.
The Curve I set to Medium Contrast and the Sharpening I set to Amount 36, Radius unchanged at 1.0, Detail unchanged at 25, and Masking to 52. Luminance I set to 8 for Noise Reduction, and Color to 15.
In the HSL panel, I raised the Saturation of the Oranges to +14, and the Blues to +5.
I decreased the Luminance value for Red to -8, raised the Orange to +14, the Yellow to +8 and dropped the Blue to -28. I chose these values by watching the image as I altered them. I knew I wanted to keep the reds and the oranges in the bricks without harming the color balance in the sky. Too much....
I skipped Split Toning, and in Lens Corrections chose the Lens profile for the EOS 24-105. I left the three sliders to their default values of 100.
As seen here
And I was basically done, except to open the image in PS and save the file unaltered as an sRGB 8bit jpg.
Here is the result. My sky is not quite as dense as John's and my clouds have less detail in them, but I think my buildings have lower noise values ( and remembering that this was a jpg, not a RAW file ) I think that is notable. I have more color in the shaded building portion in my edit.
Clearly John has captured better cloud detail, which I have not done, but I think I prefer the color and less greyish quality of my buildings.
I had the easier task as I had John's image to compare to as I worked through my steps in ACR. Having his image to compare to helps quite a bit. I look forward to doing this with the RAW file in a few days.
If I had time I should have done two versions of this image, one for the buildings and one for the sky and then blended both of them together.
I should hasten to add that all of the steps I did in Adobe Camera Raw 6.4 for this jpg, can also be done in the Raw engine in LR3, as both are the same exact Raw conversion engines. The numbers might look slightly different in LR, but the same result could be obtained.
Moderator of the Technique Forum and Finishing School on Dgrin
I still tend to think of ACR as a global editor, and do not use the local adjustment tools in ACR enough.
I was in a hurry last night, too. I will remember that tip when I get the RAW file.
Thanks again.
Pretty amazing really for starting with a jpg, isn't it?
Moderator of the Technique Forum and Finishing School on Dgrin
I think we are witnessing a dramatic, but relatively unheralded, period in raw history. Consumers are expecting their raw converters to do things that were once the realm of pixel editors, and developers, by and large, are complying. This acts in a somewhat vicious cycle: as developers add more features to their converters, customers are going to want to use their pixel editors less, putting pressure to add even more features to the raw converter. I see this as a Really Good Thing, btw, as long as cpu's are able to keep up.
What will really make things interesting is when we are able to use our third party filters directly on our raw images, with no intervening TIFF file. (Science-fiction? Maybe, but Bibblelabs has had something like that for a while now ...)
Editing in meta data is certainly better, if it can meet your image's needs.
Moderator of the Technique Forum and Finishing School on Dgrin
Nice job, Jim.
I particularly like the color and detail in the Coloseum, although I was happy with a slight yellow where you took your white point. It's hard to know, but I can't say your colors are wrong. And the contrast is quite good. What I really liked about it was how much brighter you were able to make the Coloseum. The only thing I don't like (other than the sky, which you mentioned) is the palazzo at the left side of the image. I think it's over-saturated and too bright, and the result is to draw your eye there.
With these take-aways from your image, I redid mine using the same steps, but altering the parameters (which channels to blend, gamma on the false profile, curves, etc.). Here's what I got:
It's better than my original attempt, but (except for the sky and the palazzo) not as good as yours.
Keeping the sky at least interesting is the real challenge here. It's made somewhat easier by a very easy natural mask - the B channel of LAB, inverted and auto-leveled (there's another term for auto-level in CS5, but I don't know it). But even with that, the above is about the limit of my capabilities in PS. I tried to brighten the Coloseum using the inverse of that mask with several techniques, including false profile and multiply through mask, and curves, but without much success. I have an idea of doing an HDR-like thing using various settings in Camera Raw, one set optimized for the Coloseum and the other optimized for the sky. I'll report back on that.
Your first point is an excellent one, and in my mind the key to this - the availability of tools. As the tools that people actually use in PS become available in RAW converters, then the need for PS diminishes. Of course, PS has a large set of users with a correspondingly wide set of tools, even if you just look at users doing image correction/enhancement.
I'm not so sure about your second point. This seems to me a religious issue more than a practical one. I have a relatively deep background in mathematics, and I haven't seen anything convincing on it. Actually, I don't think I've seen anything deeper than what we in the computer business called a marketing description. I'm not trying to stoke the flames here, as I have no informed opinion and am willing to be convinced - I would be grateful if someone provided me a pointer to something deeper (and no, a pointer to Thomas Knoll saying that this is so doesn't meet the criteria).
Admittedly, one can always start over with their RAW file also, but Lightroom really makes it easy to create many virtual copies of an image to alter in an infinite variety of ways if one were to desire to do so.
Not a religious issue for me, but a purely practical one for this lazy fellow.
We need to remember one of the original reasons for pixel wrangling, rather than metadata editing, was that the desktop computer of 15 years ago had nowhere near the power to do meta data editing in real time, but had to be rendered after outlining the commands, whereas pixel editors, only had to deal with a single step each time and could sort of keep up real time with the user.
Folks used to enter their commands, and go out for lunch, and supper sometimes, while waiting for the rendering. My, how times have changed.
Moderator of the Technique Forum and Finishing School on Dgrin
I want to thank John for making this file available for our readers - He now has $10 invested in postage for this thread so we all owe him a beer or a coffee someday.
Moderator of the Technique Forum and Finishing School on Dgrin
Looking at the file for color balance, I think the sky will be rendered as Daylight, but the buildings will be rendered as Shade, since they were indeed in the shade.
For the first image of the sky and the clouds, I set the color balance to Daylght, pulled the Exposure slider to the right to bring the histogram to the right edge of the box with a value of +0.75, left the Recovery slider at the default of 0, and the Fill light at the default of zero, dropped the Brightness to 25, and raised the Blacks to 40. This was actually not done in a stepwise walkthrough but interactively, adjusting one, then the other, and then going back and readjusting the previous ones to capture the rendering I wanted in the sky.
These settings gave me a deep blue, with nice clouds on the left side of the image. The clouds in the center I then attacked with the Local Adjustment Brush like Marc suggested. This left the buidlings as totally black silhouettes, but that was ok, as we will come back to them in a bit. We will open this first RAW rendering as a Smart Object by holding down the Shift key while opening the file into Photoshop. We will run the Lens Proofile for it before opening it as a smart object, as we will be using the same lens profile for the buildings, and we want the two images to match geometrically.
Clicking on the blue box in the Layers Palette with the Control key held down,, and in the new dropdown menu box, clicking on New Smart Object via Copy, gives us a new layer. We will click on this layer in the little tiny black and white box at the bottom right of the tiny image icon, while holding down the Option key, and will be dumped back in the RAW converter again with our file for a new go around, this time concentrating on the Colosseum. The inital image that opens in ACR this second time is the image of the sky with the black silhouette of the buildings, but adjusting the sliders will allow us to create a new Smart Object based on the desired appearance of the for ground buildings. So here we go
We are going to choose Shade for the color blance of the building, as we know it is not in the sunlight, but in the shade.
The Exposure slider is going all the way to +3.0, and the Blacks slider is returned to 2 or 3. The sky has gone completely and utterly white, but we already have the sky portion of the image captured in Photoshop, so no problem with the sky. I am opening Fill Light to 34 to open the shadows in the doorways, and arch sills, with the Blacks raised a bit more to 6. Recovery I left at 0.
I notice this image of the buildings is much more orange than my first attempt form the jpg.
Clarity I am raising to 25, and Vibrance to 46. Curve is left at Medium Contrast, and the Sharpening is done as Amount
65, Radius 1.0, Detail 25, and Masking all the way to 70. Noise Reduction is done with the Luminance slider to +20 and the Color slider to 48.
Skipping along to the Lens Profile I again choose to use the profile for the EOS 24-105 as I did for the first image from this RAW file. and I am done in ACR and can reopen this image into Photoshop as a new layer containing the buildings, but no sky detail. I just click OK in ACR.
Back in Photoshop I now have two layers, the first with the rendering for the sky, and the second, upper layer layer for the buildings in the forground, but not for the sky. How do I blend these? I could create an adjustment mask layer and use a brush to paint one in on the other, but I think in this image there is an easier way - How about Blend IF options as in this dialogue box obtained by clicking on the upper right are of the Layers palette column. There is almost no light blue in the building and a lot of light blue in the sky, so we can use the blue channel to blend these two images as shown in the following BLend IF dialogue box.
Now we have our image without any masking what so ever by hand. Let's type ctrl+option+shift+E to create a new third layer that incorporates everything we have done up to now in the image. Let's use the Spot healing brush to clean up the sensor dust bunnies I forgot to clean up in RAW conversion earlier.
At this point I would run the file through NoiseWare in CS4, but NoiseWare is not available in 64bit mode for CS5, so I will use DeNoise 5 from Topaz Labs as a plug in in my CS5 which is 64 bit compatible.
Next, I sharpened with Unsharp mask for Local Contrast Enhancement with amount 24, radius 50, threshold 2 and faded it with Luminosity BLending via Edit>Fade>Luminosity Blending to about 92%.
I desaturated the orange building in the lower left to satisfy John's desire for it to be less obtrusive, and darkened it a bit with Overlay Blending with a black Brush at about 12%.
And this is the final result. I thought about steepening the curve in the sky, but decided not to.
I have not previously used different color balances for separate Smart Objects, but I think it worked ok here.
Here is the final image
Is this better than the image from a jpg? The sky is certainly better.
This image was quite under exposed for the building, we had to open up about 3 fstops, and this comes with a price in noise, but let me know what you think about the final image.
Moderator of the Technique Forum and Finishing School on Dgrin
It took some time to wrap my mind around your blend-if and understand why it worked. Your observation that the buildings had no blue in them (correct) would have driven me directly to LAB. Not necessarily a mask, but possibly a blend-if on the B channel. Because Blue channel values don't really correspond to blue, I couldn't at first understand how this worked for you.
When I replicated it, I understood. The key characteristic of the sky is that the Blue channel is light, both in the blue areas and the neutral (white) areas. The Coloseum is darker everywhere in the Blue channel than any of these. This has nothing to do with whether these areas are blue or not.
This is a useful insight. I am currently exploring a similar approach to yours, but doing less in Camera Raw (just exposure, and by the way, we agree that the Coloseum is underexposed by 3 stops). My initial attempt to use a mask based on the B channel in LAB didn't work very well, as the clouds in the sky are neutral. I was thinking of blending in the L channel into the mask, but I think you've found the key to merging the two parts of this image.
For the purposes of this discussion, it seems so far that PS is necessary to do the merge of the two images.
So I used the blue channel in Blend if, and as John points out, it is not that there is no blue in the shadows, but that the blue is a much darker hue than in the sky that allows this to work quite nicely in this image.
The Blend If option is hidden in the drop down dialogue box obtained by clicking as shown in this image of the Layers palette.
Clicking on Blending Options in the drop down dialogue box brings up the previously displayed dialogue box
and as you can see, with the background layer the sky and the building silhouette, and the correct building image on the 2nd, upper layer, the blend if command passed passed all the blue from the lower layer, except the very darkest tones.
I learned Blend IF from rutt and the discussions in the book reviews about Dan's books.
How about an explanation of false gamma profiles for those of us who do not use them routinely.
It is interesting that the Colosseum was under exposed three stops, but that is exactly the difference sunny 16 predicts between sunlit objects and objects in shade.
I think the exposure was correct for the sunlight clouds in the sky.
Moderator of the Technique Forum and Finishing School on Dgrin
I also agree, Marc, but I did find more noise in the building from the dng file than I expected, but I guess when you lighten things up three stops, at ISO 400 even with a 5DMkii, you have to expect some serious noise. A little DeNoise 5 helped a lot. I think NoiseWare might have helped even more in Landscape mode.
Moderator of the Technique Forum and Finishing School on Dgrin
Maybe I don't know how to do it well yet in ACR. I find it can help, and I "guestimate" it reduces noise in my hands by maybe 60 %, but nowhere near NoiseWare does for me.
I have just been doing some scans of posters printed in a magazine with VueScan. DeNoise 5 in CS5 reduces noise, but really does not descreen a printed image of a scan. NoiseWare recognizes the screen print, and on default just cuts it right out.
I really do wish we would get a 64 bit version of NoiseWare, but from what I have heard, that is not likely. Which will keep me going back and forth between CS4 and CS5.
And I stand corrected, I see that the second image of the Colosseum is indeed ISO 400, I had it in my head that it was ISO 100 like the first, but I did not check and verify. I will correct my earlier post and that does explain the noise better in the image. Thank You for bringing this to my attention.
I do agree with doing noise reduction early in the editing workflow as well.
Moderator of the Technique Forum and Finishing School on Dgrin
I need to take some time and go over your steps in detail. Your results are great.
Please take my totally dog poo attempt down.
In the future I will remember to include the steps i took.
Sam
Or try the dng file that is available here -https://public.me.com/path_finder
Think on it, and if you really want your post pulled, I will comply.
Moderator of the Technique Forum and Finishing School on Dgrin
Where Blend-if works, it's great. It doesn't always work, unfortunately. I didn't think of using it here, but you found that it worked well.
In addition to problems like this, I use it to correct what channel blending sometimes does to the reds (dulls them). A Blend-if eliminating the blend in the dark parts of the Green channel corrects this.
I'll do that. It will take a few days, and I'll search the forum for other threads on this, possibly related to Dan's last book. The question is how deep to go, because this can take you quite deep into color spaces. But I think there's a way to discuss it at a very practical level, without all of the theory. And, in fact, that's how you use it, without thinking much or at all about the theory.
Yes, that's how I metered it, for better or for worse. I think for better, as I really wanted some sky drama in the shot.
Shot taken at ISO400, F/22 and 1/100s
Since a Canon camera was used, it would actually have been a lot easier (less noisy) if you shot it at ISO800 and the same settings - and you still wouldn't have clipped anything in the raw file....
-Or you could have shot it at 1/50s F/22 (or 1/100s F/16) in stead. Still no clipping in the raw.
This is what the image looks like in linear gamma, in the raw file.
........................................
Not that it matter for this exercise really, just FYI... :-)
I chose a few different paths when PP'ing this shot, but I imposed a few restrictions on myself to keep the workflow explainable.
*NO manual editing - this means no manually painted masks and so on. Only global adjustments
*NO Lab except if it really makes a difference
This is what I arrived at after using those restrictions - I'll try to get a step-by-step explanation of what I've done tonight, but I have to run now. Thought I'd have more time, but family activities awaits... :-)
Anyway, this is the finished result. I was going for a natural look, and I see now that I've left a bit of haloing around the leftmost part of the Colosseum/sky edge. I left the picture at a stage where I would think it's quite easy to continue working on a "personalized" look on the end result.
Thanks for the info.
I am a great fan of global adjustments. LAB is just another tool, and some things are just easier and faster there, even if you could get the same result ini RGB. But I agree that it makes it more difficult to explain, as you can't assume a general LAB understanding.
I'm quite interested in your steps.