Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: Using a polarizer with a digital camera  (Read 41911 times)
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #20 on: September 03, 2006, 04:18:51 AM »
ReplyReply

Quote
In my case, processing a Leica RAW file in ACR does improve some qualities of a picture (mainly contrast), but overall quality is visibly degraded in comparison to shots taken without a polarizer.
[a href=\"index.php?act=findpost&pid=75347\"][{POST_SNAPBACK}][/a]

Could you please make it visible to *us*; that's the only way you're going to get any real help.  A picture is worth 1000 words; a RAW file is worth 5000 words.
Logged
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #21 on: September 03, 2006, 04:27:49 AM »
ReplyReply

Quote
Could possibly be a problem where the Red channel registers lower than expected due to the polarizer, introducing artifacts during debayering because the expected levels for cross channel reproduction are not met.
[a href=\"index.php?act=findpost&pid=75322\"][{POST_SNAPBACK}][/a]

I don't think there's any kind of threshold effect; just a lack of signal relative to noise in the red channel (and in the green, but to a lesser extent).  The red channel gets multiplied by almost 2 to achieve daylight WB - after WB the readout noise in the red channel is higher at ISO 100 than it is in the green channel at ISO 800.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #22 on: September 03, 2006, 08:42:17 AM »
ReplyReply

Quote
I don't think there's any kind of threshold effect; just a lack of signal relative to noise in the red channel (and in the green, but to a lesser extent).  The red channel gets multiplied by almost 2 to achieve daylight WB - after WB the readout noise in the red channel is higher at ISO 100 than it is in the green channel at ISO 800.
[a href=\"index.php?act=findpost&pid=75355\"][{POST_SNAPBACK}][/a]

That probably explains the relatively large blobs of red signal differences, but does that also explain the overal grainy noise? This image doesn't look like ISO 100 at all, it's that bad. I'm thinking more along the lines of how a polarizer can change the appearance of colors significantly enough so that the default debayering co-efficients per color aren't valid.

Obviously, this is something that should easily be reproducable, but unfortunately the weather here is completely overcast and gray. I'm still going to do some tests though, because it may help identify the real culprit. If for example the UV content is throwing the capture off, it should not be so in conditions lacking UV...
Logged

Regards,
Oscar Rysdyk
theimagingfactory
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #23 on: September 03, 2006, 09:54:30 PM »
ReplyReply

Quote
That probably explains the relatively large blobs of red signal differences, but does that also explain the overal grainy noise? This image doesn't look like ISO 100 at all, it's that bad. I'm thinking more along the lines of how a polarizer can change the appearance of colors significantly enough so that the default debayering co-efficients per color aren't valid.

There's an easy way to test that theory; see what the image looks like without demosaicing.  Can you put the RAW file somewhere to download it?

Quote
Obviously, this is something that should easily be reproducable, but unfortunately the weather here is completely overcast and gray. I'm still going to do some tests though, because it may help identify the real culprit. If for example the UV content is throwing the capture off, it should not be so in conditions lacking UV...
[a href=\"index.php?act=findpost&pid=75362\"][{POST_SNAPBACK}][/a]

Take RAW images with different combinations filters (CP, UV, CP+UV, none) in manual mode with the same settings on a tripod, exposed so that the one with no filters is exposed to the right.  Then we can see what the difference is between them.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #24 on: September 04, 2006, 07:17:46 AM »
ReplyReply

Well, here is a first preliminary test.

Canon DR (300D) Aperture priority @ f11

Artificial daylight @ ~D55

B+W Linear polarizer

Battery happened to be completely charged prior to this test, but earlier tests with depleted battery showed no significant differences.

Following image shows the cccard with 3 exposures:
1. Incorrectly polarized, camera recommended exposure (2s)
2. Incorrectly polarized, recommended exp + 1stop (4s)
3. Correctly polarized, camera recommended exposure (4s)



Incorrect polarization = correct pol + 90degr rotation

For incorrect polarization, +1stop results in the tiniest over-saturation, compensated for in ACR by -0.10 exposure correction.

For correct polarization, camera recommended exposure results happened to be perfect ETTR for sRGB.

Clearly, the correctly polarized result is more saturated and has more contrast.


Following image compares the blue patch of the +1stop incorrectly polarized image vs the correctly polarized image @ 200% magnification.




Given the significant difference that polarization makes in color rendition, even in this situation, but a complete lack of noise differences, I strongly suspect that I can not reproduce the artifacts with my equipment. It might be of course that a (B+W) linear polarizer has a completely different interaction than a circular polarizer, or something is significantly different in outside conditions.

RAW files:
Correctly Polarized
Incorrectly Polarized
Incorrectly Polarized (+1stop)
Logged

Regards,
Oscar Rysdyk
theimagingfactory
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #25 on: September 05, 2006, 11:58:06 AM »
ReplyReply

I took the liberty of examining the EXIF data of the noisy shot of the building/sky, and the problem is simply underexposure. The Exposure setting used in ACR was +2.00. With Canon DSLRs, you should expose in-camera so that ACR's exposure setting is between 0 and -0.35 during conversion. If you have to adjust below -0.35, you're going to start having some clipped highlights, and if you have to adjust above 0, you're going to have elevated noise levels. In this instance, the presence of a polarizer is irrelevant; the real problem is operator error.
Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8849


« Reply #26 on: September 05, 2006, 10:10:53 PM »
ReplyReply

Quote
In this instance, the presence of a polarizer is irrelevant; the real problem is operator error.
[a href=\"index.php?act=findpost&pid=75584\"][{POST_SNAPBACK}][/a]

Indeed! Sometimes the most obvious reason is the correct one. A 1/400th sec at ISO 100 and f10??

In certain bright and contrasty situations, the dynamic range of Canon DSLRs is not great enough. Some of the first shots I took with my new 5D and 24-105 IS lens were of a sunset. I was actually looking for the flare problem in the lens, that everyone was talking about at the time. Exposing for the brightest parts of the image (these were not direct shots into the sun), I discovered the shadows were disturbingly noisy in a way that was impossible to remove, like a fixed irregular pattern like coarse jute or burlap.

But the OP's shot in this thread is just plain underexposed.
Logged
iconoclast
Newbie
*
Offline Offline

Posts: 14


« Reply #27 on: September 06, 2006, 01:19:22 AM »
ReplyReply

We seem to be homing in on the essence of this problem. Everything points to the underexposure with polarizers.

Since several of us had a similar problem with different circular polarizers and digital cameras of different makes and models, is it safe to say that to get proper exposure in those circumstances, we should override the in-camera meter by +1.5 to +2 EVs? Any other conclusions?
Logged

Regards,
Rob
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #28 on: September 06, 2006, 10:23:23 AM »
ReplyReply

Quote
We seem to be homing in on the essence of this problem. Everything points to the underexposure with polarizers.

Since several of us had a similar problem with different circular polarizers and digital cameras of different makes and models, is it safe to say that to get proper exposure in those circumstances, we should override the in-camera meter by +1.5 to +2 EVs? Any other conclusions?

One should always review the histogram after shooting to verify proper exposure, regardless of whether a polarizer is used or not. I'm skeptical of the notion that a polarizer consistently causes underexposure; I've never experienced such a problem. But I use circular polarizers; linear polarizers can cause problems with metering and autofocus and should never be used on SLR cameras, film or digital. But if one uses a circular polarizer, there's no reason to always dial in +2 EC.
Logged

John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #29 on: September 06, 2006, 01:47:24 PM »
ReplyReply

Quote
For correct polarization, camera recommended exposure results happened to be perfect ETTR for sRGB.

Actually. CRW_6081 is actually more than a stop away from any RAW clipping.  The brightest RAW pixel is about 1900, out of 3967.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #30 on: September 06, 2006, 03:35:52 PM »
ReplyReply

Quote
Actually. CRW_6081 is actually more than a stop away from any RAW clipping.  The brightest RAW pixel is about 1900, out of 3967.
[a href=\"index.php?act=findpost&pid=75713\"][{POST_SNAPBACK}][/a]

Yes, my apologies, I suppose "perfect ETTR" is a misnomer in this case. ACR in ProPhotoRGB suggests there is still another 1.3 stops to go. In sensor space it may be even more.

I only shot upto +1 stops for both cases, and the brightest incorrectly polarized image matched the normal exposure of the correctly polarized image. The results were already conclusive in that there is absolutely no reason to believe that I can replicate the problem...
Logged

Regards,
Oscar Rysdyk
theimagingfactory
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #31 on: September 06, 2006, 03:55:12 PM »
ReplyReply

Quote
Yes, my apologies, I suppose "perfect ETTR" is a misnomer in this case. ACR in ProPhotoRGB suggests there is still another 1.3 stops to go. In sensor space it may be even more. [a href=\"index.php?act=findpost&pid=75720\"][{POST_SNAPBACK}][/a]

ACR is not mathematically accurate with the exposure slider.  There is no real exposure bias control in ACR.  There should be one that biases the RAW data in a completely linear fashion, and that one should be called "Exposure".  ACR's "Exposure" is a combination of exposure, and built-in curves.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #32 on: September 06, 2006, 04:11:37 PM »
ReplyReply

Quote
ACR is not mathematically accurate with the exposure slider.  There is no real exposure bias control in ACR.  There should be one that biases the RAW data in a completely linear fashion, and that one should be called "Exposure".  ACR's "Exposure" is a combination of exposure, and built-in curves.
[a href=\"index.php?act=findpost&pid=75721\"][{POST_SNAPBACK}][/a]

Well, ACR internals are supposed to be linear space, so the exposure slider may well be mathematically accurate. The histogram however, only shows colorspace converted data.
Logged

Regards,
Oscar Rysdyk
theimagingfactory
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #33 on: September 06, 2006, 08:30:40 PM »
ReplyReply

Quote
Well, ACR internals are supposed to be linear space, so the exposure slider may well be mathematically accurate. The histogram however, only shows colorspace converted data.
[a href=\"index.php?act=findpost&pid=75724\"][{POST_SNAPBACK}][/a]

Shoot a greystep card at -4, -2, 0, +2, and +4 EC, and then do the opposite with them in ACR.  Very different renditions.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8849


« Reply #34 on: September 06, 2006, 09:40:45 PM »
ReplyReply

Perhaps Iconoclast is trying too hard to be an iconoclast here. His results make no sense to me. I just tried a Hoya circular polarizer on my Sigma 15-30, at 30mm with my 5D. It's a bright, sunny day. With camera in aperture priority mode and meter in pattern mode (evaluative), pointing the camera at a scene consisting of half sky and half river and fields, I got the following results at f10; with polariser, 1/40th to 1/80th depending upon the circular adjustment of the filter; without polariser, camera pointed at the same scene, 1/200th sec.

I see no reason why the camera's metering system would not read a correct exposure with the addition of a circular polariser.

I once met a guy on my travels who was using a 10D with a polariser permanently fixed to his zoom lens. He'd filled up all his flash cards and had nowhere to store the images, so I downloaded them on my laptop and burned them to CD for him. During the process, I noticed that a lot of his images were not sharp. His shutter speeds were too slow. I explained to him that the polariser was causing this and that it wasn't a good idea to have the polariser permanently fixed to his lens.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #35 on: September 07, 2006, 12:07:41 AM »
ReplyReply

Quote
Shoot a greystep card at -4, -2, 0, +2, and +4 EC, and then do the opposite with them in ACR.  Very different renditions.
[a href=\"index.php?act=findpost&pid=75741\"][{POST_SNAPBACK}][/a]

Yes, but that is to be expected because that is related to the tonal curve (which will be different for any RAW converter).

The clipping point however should react completely predictable. If it takes a +1.3 exposure correction in ACR to clip, then a +1.3 exposure bias in camera capture should exactly move the data to the point of clipping.

Problem of course is that bright colors may commonly clip sooner than neutral highlights. In addition, the detail tab settings will SIGNIFICANTLY change the histogram as wel. So you may see clipping in the ProPhoto histogram and still have RAW data latitude. Only a RAW data histogram (possibly rendered in "stops" for perceptual uniformity) would show the true ETTR latitude.
Logged

Regards,
Oscar Rysdyk
theimagingfactory
John Sheehy
Sr. Member
****
Offline Offline

Posts: 838


« Reply #36 on: September 07, 2006, 08:07:26 AM »
ReplyReply

Quote
Yes, but that is to be expected because that is related to the tonal curve (which will be different for any RAW converter).
[a href=\"index.php?act=findpost&pid=75759\"][{POST_SNAPBACK}][/a]

Obviously, clipping points will be different, but unclipped, rendered tones will be different, too, which proves beyond a doubt that there is no simple linear adjustment to the data with the "exposure" slider.  ACR does not allow you to expose all the way to the right, and get a reasonable render with negative exposure compensation.  You need another converter for that.

Like everything Adobe, ACR has some great features, and some big, gaping holes.
Logged
Rob C
Sr. Member
****
Offline Offline

Posts: 12215


« Reply #37 on: September 07, 2006, 01:17:45 PM »
ReplyReply

iconoclast

May I go back to your very first post in this thread where you write that you went down to the beach with a girl to shoot some summer pictures. Now, I hope that the girl was only along for the ride - the use of a polarizing filter on flesh is not a pretty effect at all, certainly on colour film, where the effect is that of turning tan into brickwork.

If you take the time to look carefully at some of the better model photographers who do this kind of work, you will not see many of their pictures where the highlights have been polarized off the skin! On the contrary: the industry makes great use of the application of makeup and sun-oil products to create, purposely, a noticeable shine. There are, of course, as with everything else, exceptional cases, as in some fashion work where the whole deal is about creating saturated colour in fabric or even location. This can be harsh as hell, but if the effect is satisfying to the commissioning editor...

You can, of course, use it (the filter) with success if you keep the figure small in the frame and make the location the principal feature, particularly if you can manage to get the beautiful fluffy white clouds that some places seem to create - then, the polarizer will do its work well and be worth its weight in gold.

As with so much else: use with caution!

Best wishes - Rob C
Logged

iconoclast
Newbie
*
Offline Offline

Posts: 14


« Reply #38 on: September 08, 2006, 01:07:56 AM »
ReplyReply

Quote
iconoclast

May I go back to your very first post in this thread where you write that you went down to the beach with a girl to shoot some summer pictures. Now, I hope that the girl was only along for the ride - the use of a polarizing filter on flesh is not a pretty effect at all, certainly on colour film, where the effect is that of turning tan into brickwork.

If you take the time to look carefully at some of the better model photographers who do this kind of work, you will not see many of their pictures where the highlights have been polarized off the skin! On the contrary: the industry makes great use of the application of makeup and sun-oil products to create, purposely, a noticeable shine. There are, of course, as with everything else, exceptional cases, as in some fashion work where the whole deal is about creating saturated colour in fabric or even location. This can be harsh as hell, but if the effect is satisfying to the commissioning editor...

You can, of course, use it (the filter) with success if you keep the figure small in the frame and make the location the principal feature, particularly if you can manage to get the beautiful fluffy white clouds that some places seem to create - then, the polarizer will do its work well and be worth its weight in gold.

As with so much else: use with caution!

Best wishes - Rob C
[a href=\"index.php?act=findpost&pid=75800\"][{POST_SNAPBACK}][/a]
Thank you, Bob C.
Logged

Regards,
Rob
Hank
Sr. Member
****
Offline Offline

Posts: 679


« Reply #39 on: September 08, 2006, 09:08:51 AM »
ReplyReply

FWIW, I use polas (B/W schneider and Hoya) and don't see anything resembling the problems you folks are discussing.  To avoid brand wars, I'll only say that I'm not using  Canon.  

Is there any possibility that all this is something specific to Canon's sensor technology?  Serious question, and flames are deflections.
Logged
Pages: « 1 [2] 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad