• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

A Problem File

I have been working this image (actually a series of images) as the sun set over the Carbon River for over month and I am still not happy. I have tried half a dozen RAW converters and still the CA is giving me grief.

SPE25921_RSE_01_demo.jpg


Sunset Over The Carbon River, Mt. Rainier National Park, USA

Artistically, being 3-5 weeks earlier to the locale may have yielded batter light (nice direct backlight down the valley) or it may have been submerged. This local is currently gone in this years flooding but it will be back once the waters recede.

You can find the problem file at:

http://www.envisagement.com/opf/SPE25921.CR2

The problem is that Canon's EF-S 18-55/4-5.6 lens has serious optical problems in the corners in contrasty light. Depending upon the RAW converter there is an issue with CA* in 2 or 3 corners of the image.

The upper left has CA in the trees against the sky at a moderate distance from the lens.

The mountain at upper right has CA on the top of the ridge above the clouds at an extreme distance from the lens.

And in some converters the contrasty edges between the dark sand/silt and reflected skylight off the water yields CA issues at a relatively short distance from the lens.

Add in the that the DR of this image pushes the limits of a RAW file on a Canon CMOS sensor and this makes a nice problem file.

I was originally shooting for HDR, but the CA makes fixing each frame time consuming. My current solution is to do 3 RAW Conversions with CA corrected for each problem area separately and then combine them with layer masks to get a good converison. Doing it for 11 shots (33 RAW conversions) seems a bit excessive and time consuming to do HDR and I was hoping people might share their expertise and see if one cannot get a single clean conversion and show how they did it.

thanks,

Sean





*CA -> Chromatic Abberations
 

Ray West

New member
Hi Sean,

On my screen, I can see what looks like ca in the image as posted here. I downloaded the cr2 image from your link, loaded into cs2, and I can't see any ca. I can see the sort of light diffraction you get though tree branches, etc., if there is a strong light behind them, but I can see no ca around the mountain tops, or wherever. This is looking in raw, up to 400%. If I just convert it into pro photo rgb (16 bits), no ca there, either.

I am puzzled, unless you are seeing jpeg artifacts, or similar, if you are converting.

Would you like to buy my glasses? I've opened it again, still can't see what you mean.

Best wishes,

Ray
 

Ray West

New member
Hi Sean,

I've just loaded it into dpp, there is ca there, visible in the raw file, at 200%, not much, but there.

Best wishes,

Ray
 
Hi Ray,

My killer is the implicit RAW conversion with Merge To HDR from Bridge is that the ridgeline at upper righ




Ray West said:
I downloaded the cr2 image from your link, loaded into cs2, and I can't see any ca. I can see the sort of light diffraction you get though tree branches, etc., if there is a strong light behind them, but I can see no ca around the mountain tops, or wherever. This is looking in raw, up to 400%.

<pause to get latest Camera RAW>
In ACR there is mild cyan fringing on the ridgeline at upper right visible at 400% zoom. I should note I am talking about the ridgeline above the line of clouds in the valley and not the darker ridgeline right in the corner. If you go to the lens tab and set the Fix Red/Cyan Fringe to -40 it goes away. Similarly set the Blue/Yellow to -25 for the trees at upper left. While this cleans up the image, the shadow detail is close to non-existent.

Hence using the implicit ACR conversion in Merge To HDR left me with a very nasty cyan halo at upper right (over 1 pixel wide) and dismal color on the trees at upper left.

I will try this again using PSDs in Merge To HDR with the corrections applied.

My main problem with ACR is the dismal rendition of shadow detail (noise where others render shape). And since I was shooting for HDR which exacerbated the CA a RAW conversion that lacked shadow detail was simply not acceptable for my vision.

I just tried this using PSDs in Merge To HDR and I still have some CA in the absolute corners (or at least artifacting). I will have to check the registration of the frames and try again when I have more time (fxing subpixel registration issues involves massive upsampling and lots of memory access).

This yields (yet one more subject color rendering):

SPE25921_ACR_HDR_16%20copy.jpg



There are still artifacts on the ridgeline above the clouds which may be PP issues but the RAW conversion is not perfect.

I will experiment some more.

Ray West said:
I've just loaded it into dpp, there is ca there, visible in the raw file, at 200%, not much, but there.

This may be a matter of taste. For me, any CA or PP artifact that will be visible in a print it way too much.

thansk,

Sean <smile>
 

StuartRae

New member
Sean,

Here's an example converted by RSP and then PP'd using Shadow Illuminator. Total time < 5 minutes.

Ir's probably a bit more extreme than I'd normally like, but it does bring out the details in the rocks and river.

Like Ray, I think the artifacts round the branches are more likely to be diffraction than CA.

SPE25921-02.jpg


Regards,

Stuart
 

John_Nevill

New member
Sean,

I tried to email you a full size jpeg last night, it got bounced. Anyhow here's the gist of it.

I can see the CA problem, it looks like differing types of CA, which conflict when you try
and remove. I also think that some of it is diffraction

I've tweaked it like mad with a mixture of CA removal using the ca tool in
SP on the trees and then used the sharpening false outline to help control
the mountain ridge.

I also used the DR expansion and a few other functions to adjust. Its
probably not right on the colour / brightness / sat as its difficult to
visualise the scene.

SDJLN.jpg


I can email the larger jpg again to if different address, let me know.
 

Ray West

New member
Hi Sean,

I just realised, you are using the 'kit lens'. My version would not give as good as you've got, or rather you got better from yours, then I got from mine ;-) I was going to suggest, if cash was tight, that the 50mm 1.8 would be much, much, better, so I loaded your image into irfan view, to look at the exif, and saw you were at 25mm focal length, so you'd have to be further away. Now, I think you said you were wanting hdr, three images, I guess this is the underexposed one. Why do you need the shadow area detail? I know you are looking for a work flow, but I honestly think you will not achieve it too well, because of the lens. I got such muddy colours with that lens.

There was a guy, I found him via fm forum, who did a sort of lens distortion correction action, for cs2, a bit like dxo. I sent him some files, and he did a profile for an older lens of mine. I do not remember if it dealt with CA, but it did what he said it would. If you can't get to what I'm on about, I'll dig back through, later on, sometime, and find the details.

Best wishes,

Ray
 
StuartRae said:
Here's an example converted by RSP and then PP'd using Shadow Illuminator. Total time < 5 minutes.

Ir's probably a bit more extreme than I'd normally like, but it does bring out the details in the rocks and river.

Like Ray, I think the artifacts round the branches are more likely to be diffraction than CA.
Hi Stuart,

Interesting example and a I agree it is a bit extreme on the shadow correction for my tastes too. The Shadow Illuminator plugin looks interesting and I played with the demo for half an hour. I may buy it even though it is single threaded (no dual core speedup) as the price is low for a useful filter.

I agree there is more than CA on the trees at upper left, but there is still some CA. What I like about this shot from a problem perspective is that there are three distances in the shots corners bringing out the worst in a low quality lens.

thanks,

Sean
 

Ray West

New member
Hi John,

I 'm fairly certain that is it - I think it came out of some panoramic software, gnu license, or something. I found it very helpful at the time, but for what I do now, I don't use it, but then, I'm a luddite... I wish canon never made that lens. Its a nice size, (nice price, too), nice working range for a crop camera, but the rest of its specification sort of sucks, whereas the little 50mm has a very good image quality.

I would be very interested to see Sean's results at using hdr, on even one set of images, because if there was any way of nailing the iq on the lens, it would be very useful.

Best wishes,

Ray
 

Ray West

New member
Hi Sean,

What size print are you going for? I take it you use qimage or similar/better. In dpp, I noticed similar ca on the top of the fir tree branches, as I mentioned in the mountain area, but compared to other problems? in the tree area, it was insignificant.

In reality, all this stuff is playing with numbers, sums, algebra, that sort of thing. It is a compromise. Show the detail of the fir trees, you show the noise. dpp sums are different than cs2, etc. How does cs2 know its ca, and not a distant forest fire, the other side of the ridge?

Best wishes,

Ray
 
John_Nevill said:
Sean,

I tried to email you a full size jpeg last night, it got bounced. Anyhow here's the gist of it.

I can see the CA problem, it looks like differing types of CA, which conflict when you try
and remove. I also think that some of it is diffraction.

I've tweaked it like mad with a mixture of CA removal using the ca tool in
SP on the trees and then used the sharpening false outline to help control
the mountain ridge.

I also used the DR expansion and a few other functions to adjust. Its
probably not right on the colour / brightness / sat as its difficult to
visualise the scene.

I can email the larger jpg again to if different address, let me know.

Hi John,

The email bouncing sounds strange. What email address did you use? opf AT my domain name (in sig)? How large was the file? The address should support up to 10 MB attachments. If you used the opf address and it failed, then please try sdemerch AT gmail DOTT com.

I agree diffraction is likely an issue with the trees at upper left but that fringing is reduced by fixing the CA nonetheless. Sadly, I cannot remember where I was focussed so it may have been 10 m or infinity.

The best rendition I have in terms of luminosity is out of DxO but that lacks all color detail in the shadows and lower midtones and the 3/4 tone and highlight colors were off from reality (but I liked them).

RAW Therapee handled the CA well, but I did not care for the noise or color.

I was okay with the color in RSE, but I did not particularly care for the CA issues. I tried RSP long ago but never used any of the extra features making RSP a bad buy. I am disappointed Pixmantec is gone because RSE is a top performer in terms of speed.

I will probably try Silkypix on the HDR and do 15 conversions (3 per shot) and see if I cannot get an HDR I like. I will have to figure out what the sharpening false outline feature is.

thanks,

Sean
 
Hi Ray,

Ray West said:
I just realised, you are using the 'kit lens'. My version would not give as good as you've got, or rather you got better from yours, then I got from mine ;-) I was going to suggest, if cash was tight, that the 50mm 1.8 would be much, much, better, so I loaded your image into irfan view, to look at the exif, and saw you were at 25mm focal length, so you'd have to be further away. Now, I think you said you were wanting hdr, three images, I guess this is the underexposed one. Why do you need the shadow area detail? I know you are looking for a work flow, but I honestly think you will not achieve it too well, because of the lens. I got such muddy colours with that lens.

I also have the 50/1.4 and 100/2.8 macro. The problem was, I was in the middle of a river bed in between two channels of the river and there was mist everywhere (it is a rain forest) so there was no way I was changing lenses. Heck, after working the shot (40 minutes of watching the light shift and exploring to choose a composition plus 40 minutes of hunched over shooting as the sun set) and at the end I quit shooting because my lens was fogged up and I was cold. I got back to the trailhead and found that not only had the dew/mist settled on the lens but that I too was coated with dew/mist (and hence cold).

The composition itself would never have worked with a wider lens as I was about 0.6 m below a sand/silt bar right at the edge of the water and the sand bar would have eclipsed the water in the foreground for the shapes and curves I wanted. Longer glass would have also reduced the wide angle distortion and changed the curves of the image (I wander up and down a 0.5 km stretch of the river to find the composition I wanted).

As to the kit lens, I rather dislike it (it is not enjoyable to shoot with) but with what I shoot and the fact that close to home landscapes are better at 200-300 mm (which I do not have a lens for) due to power lines and the fact that compression of perspective will make distant mountains loom larger. Hence investing in wider glass is very low priority as will almost never use it (not for local landscapes, not for people, and not for macros).

I will likely pick up the Sigma 12-24 for my next trip to a national park where wide angles work well as I found 18 mm to not be wide enough for some compositions. But that is a low priority.

For HDR I have 5 shots for this set (I cannot remember the last time I used less 5 shots) and I have used up to 13 in more contrasty light. The problem though is HDR is very sensitive to lens aberrations and throws out very nasty colors.

As for shadow detail, what I really want is soft contrast on the rocks in the riverbed and a hint of shape in the silhouette at left with some distinction between depths at the treeline and ridgeline in the shadows at right. I like soft mildly textured shadows rather than strong clipped shadows.

Ray West said:
There was a guy, I found him via fm forum, who did a sort of lens distortion correction action, for cs2, a bit like dxo.


I suspect this is PTLens which I have used in the past. It is okay, but not particularly helpful for a scene like this (no dominant straight lines) and it does not handle CA at all. Plus CA is probably better handled before lens projection corrections.

Ray West said:
What size print are you going for? I take it you use qimage or similar/better. In dpp, I noticed similar ca on the top of the fir tree branches, as I mentioned in the mountain area, but compared to other problems? in the tree area, it was insignificant.

This will probably never be printed larger than 8x12 inches as I want the shot with backlight flowing down the valley which is more of a mid-August than late September image. I am more interested in simply getting the shot to show my vision and I in general think the composition is okay. I would like to get the image of a quality (accepting the fringing on the trees at upper left) that would print an adequate 16x24 inch print. I have been working this image more as a PP challenge than with panic over getting a print as the basic scene is now on my 5 year plan for an August sunset (sunrise fails due to the valley pointing WNW. For anything over 12x18 inches I would likely hand it off to the printer for increasing resolution. Up to 12x18 inches I just print them at the local drug store off a Fuji Frontier and put them to 300 DPI in PS. Worst case, I can always add noise to the print to hide artifacts.

In the end though, I am simply vexed with the composition and I want to get it to work as HDR but need a clean RAW conversion to get there.

Ray West said:
In reality, all this stuff is playing with numbers, sums, algebra, that sort of thing. It is a compromise. Show the detail of the fir trees, you show the noise. dpp sums are different than cs2, etc. How does cs2 know its ca, and not a distant forest fire, the other side of the ridge?

That I cannot answer. All I know is what my vision is as reality is already forgotten. The sky was cyan, the clouds pink, and there was clear detail in the soft light over the riverbed. Artistically, I personally do not like clipped black silhouettes and favor a dash of shape without detail and a hint of color.

thanks,

Sean
 

Tim Gray

New member
So here's my effort, as one of the previous posters mentioned there is not really much CA.

Here are 3 images:

converted with Lightroom, not processed

converted.jpg


100% crop

crop.jpg


processed using my normal workflow - without noise reduction - given how much I've opened the shadows, I'd probably run it through Neat Image.

processed.jpg


Conversion was done in Lightroom

Exp 0
Recovery 34
Fill 43
Blacks 0
Brightness +45
Contrast 0
Highlights +21
Red Luminosity -75 (you could fuss with this if you wanted more vibrant reds in the sky)

I find that Lightroom has all the ooompf needed to adjust a (single) HDR file to deal with clipping without resorting to the process twice then merge technique. Which makes sense, if the data is available in the single RAW file, why should you need to process twice.
 
Tim Gray said:
I find that Lightroom has all the ooompf needed to adjust a (single) HDR file to deal with clipping without resorting to the process twice then merge technique. Which makes sense, if the data is available in the single RAW file, why should you need to process twice.

The need to process twice is IMHO because you need radically different shadow contrast and highlight contrast, and even locally adaptive tonemapping.

This is my rendition, based on 2 conversions in RawShooter Premium (automatically removed most of the lateral CA):
SPE25921_2xRSP+Tonemap.jpg
SPE25921_2xRSP+Tonemap.jpg

One conversion for not clipping the highlights, another conversion of +3 stops EC to open the shadows with more detail and saturation. I also used some color noise reduction and a tint correction on the 'shadow' conversion.

I chose a more dramatic level of lighting, but you'll see there is more shadow detail in the trees and rocks in the foreground, and in the sun lit clouds.

My postprocessing method was:
- Blend the two conversions,
- Promote the 16-bit result to 32 bit/channel, and back to 16-b/ch, which allowed to adaptively tonemap the result,
- do some additional curves adjustments (because CS2's tonemapping is so poorly implemented),
- add some overall saturation, and reduce some blue saturation,
- convert to sRGB (which introduced some clipping),
- use the Shadow and Highlight control for zero clipping and a 1% highlight adjustment to counteract the clipping.

There are some blending artifacts in the tree tops, which could have been reduced with some more work. However, I think this subject justifies a real HDR approach. That would allow to better preserve the haze which adds to the perspective and atmosphere, and reduce the shadow noise.

For the reduction of CA, which is quite visible in the tree tops and the mountain crest, I usually get decent reduction from RSP, or from the ACR controls. DxO does a much better job, and is tunable, but I'm not impressed with the output resolution (the lack of which does avoid moire artifacts). One can also use PanoTools based Photoshop CS2 scripts from PTShift, but I find that route too slow to setup the corrections for a lens (the results can be reused though, so it might be useful if you build a database of lenses and apertures).

I'm still amazed at the substandard methods for CA removal in many/most software/raw converter packages. It's a process that is ideally suited for intelligent locally adaptive corrections. Maybe the RawShooter input to Adobe will make the future a little brighter.

Bart
 
Top