Reading mode: Light Dark

To Joe: software v optics (re - weak AA)

Go to last post
Great Bustard said:
Let me summarize, since you appear to have missed it the first time around: DPR uses the same processing for all systems which is likely not the best approach for anyone seeking optimal results.
Click to expand...
Your statement confuses me. I and others have previously quoted and linked to the DPR description of test methodology in which they specifically state they apply USM tuned to the camera in resolution tests, and go on to state what the typical USM values are. They also specifically state in the E-3 review that they tried various sharpening approaches to attempt to extract more resolution, but it didn't improve the result.

How does that constitute using the same processing for all systems which is likely not the best approach for anyone seeking optimal results ? They specifcally describe how they use the same methodology but make USM adjustments for the specific camera, and go further to describe how they did try to get optimal results on the E-3. What is unclear about that?
--
Bob Cole
http://www.robertcolephotography.com
 
ROC124 said:
Great Bustard said:
ROC124 said:
ROC124 said:
Show me the proof. And it had better not be MTF because MTF measure contrast and contrast adjustments will alter the MTF as DPReview has already noted.
Click to expand...
Not accepting MTF as proof of resolution is like not accepting a mathematical proof that 2 times 6 equals 12. How can you NOT accept industry standard resolution tests as proof of resolution found in those tests? Again, the results are documented on this site. You can certainly disagree with the conclusion that the AA filter had anything to do with the results, but you are in no position to demand proof of a rational conclusion when you can't offer any proof that the conclusion was wrong.
Click to expand...
This is central to the whole discussion. A manufacturer MTF chart for a lens measures contrast at a particular resolution (lp/mm on the sensor). A lens (system) test measures resolution (lw/ph) at 50% contrast.

So, the point Tim is making is that since the processing can affect the contrast, it will also affect the MTF-50 score. In addition, he is making the assertion that applying the same processing to both systems is not optimal, especially if they have different AA filters (and pixel counts).

This is the exact same reason the DxOMark has different measures for DR than IMATEST. DxOMark measures the DR of the unprocessed RAW file, whereas IMATEST measures the DR of a processed photo. And, yeah, the differences can be huge .

So, Tim is rejecting the MTF-50 results on this basis. I have to say that this is a very valid objection, and the reason I am so interested in this discussion. I've always been under the impression that MTF-50 scores were comparable across systems, but now I'm thinking that differences in processing make that a problem, and a problem worth sorting out.
Click to expand...
I'm aware of all those issues and the complications of comparing tests with different methodologies and input assumptions. Nature of the beast, but it doesn't invalidate any of them, just makes direct comparisons more difficult. So what? Seems like there has already been general agreement that the differences often don't matter much in practical application.
Click to expand...
What Tim is asking, and what I am wondering, is that if the files were "optimally processed" for each system, rather than the same processing being applied by both, how much of a difference is really there?

With a really sharp lens, the difference between the E3 and E5 would be 10%. However, some MTF-50 tests show a 30% differential. What Tim is questioning is the "validity" of the 30% figure.

The point he's trying to make, I think, is that the "actual" difference is a lot closer to 10% than 30%, and the risk of artifacts isn't worth the small gain he feels the weaker AA filter gives.
ROC124 said:
One of the original justifications for this protracted discussion is that some here claim substantial, visible resolution differences between the E-3 and E-5, with the speculation that the AA filter is largely but not enitirely responsible. Others say they don't see substantial differences. Big deal, humans are prone to hyperbole in both directions. There isn't any standard for "substantial" and it doesn't matter what test numbers show - if you don't perceive the differences you aren't inclined to believe the tests. That really isn't a technical issue, it is a human issue, and no technical proof is likely to be satisfactory.
Click to expand...
Well, of course no "big deal" -- they're just cameras! ;) But it is interesting to know just how much more detail the weaker AA filter gives over a stronger AA filter (that's basically the bottom line in these discussions) and if that additional detail is "worth" the risk of artifacts, sometimes rather severe.

Of course, that's entirely subjective. But Tim feels that if reviews are comparing identically processed E3 and E5 files, it gives a false impression as to what the real differences are for a photographer who competently processes their files.
 
ROC124 said:
Great Bustard said:
Let me summarize, since you appear to have missed it the first time around: DPR uses the same processing for all systems which is likely not the best approach for anyone seeking optimal results.
Click to expand...
Your statement confuses me. I and others have previously quoted and linked to the DPR description of test methodology in which they specifically state they apply USM tuned to the camera in resolution tests, and go on to state what the typical USM values are. They also specifically state in the E-3 review that they tried various sharpening approaches to attempt to extract more resolution, but it didn't improve the result.

How does that constitute using the same processing for all systems which is likely not the best approach for anyone seeking optimal results ? They specifcally describe how they use the same methodology but make USM adjustments for the specific camera, and go further to describe how they did try to get optimal results on the E-3. What is unclear about that?
Click to expand...
For the 12 MP E5, DPR said:

http://www.dpreview.com/reviews/olympuse5/page6.asp

On this page we're looking at both JPEG and Raw resolution. For a (more) level playing field we convert the latter using Adobe Camera Raw. Because Adobe Camera Raw applies different levels of sharpening to different cameras (this confirmed) we use the following workflow for these conversions:
  • Load RAW file into Adobe Camera RAW (Auto mode disabled)
  • Set Sharpness to 0 (all other settings default)
  • Open file to Photoshop
  • Apply a Unsharp mask tuned to the camera, usually 100%, Radius 0.6, Threshold 0
  • Save as a TIFF (for cropping) and as a JPEG quality 11 for download
For the 24 MP NEX 7:

http://www.dpreview.com/reviews/sonynex7/page18.asp

DPR said the exact same thing. Now, even if they were typical values, given the radical differences between cameras, it does raise the question as to why there would be a typical value at all, as if sharpening doesn't require much tuning to any particular camera at all.

The entire point Tim was making is that sharpening made a marked difference between the E3 photo and E5 photo, and that he used rather different values for each.

So, combine that with your saying that DPR "tried various sharpening approaches to attempt to extract more resolution, but it didn't improve the result", then this is very curious indeed. The implication is that sharpening doesn't affect the MTF-50 score, which doesn't ring as true at all.
 
Great Bustard said:
What Tim is asking, and what I am wondering, is that if the files were "optimally processed" for each system, rather than the same processing being applied by both, how much of a difference is really there?
Click to expand...
Only Olympus knows for sure, and I'm quite sure they won't disclose that info.
Great Bustard said:
With a really sharp lens, the difference between the E3 and E5 would be 10%. However, some MTF-50 tests show a 30% differential. What Tim is questioning is the "validity" of the 30% figure.

The point he's trying to make, I think, is that the "actual" difference is a lot closer to 10% than 30%, and the risk of artifacts isn't worth the small gain he feels the weaker AA filter gives.
Click to expand...
All that is fine, but again it is subjective whether it is worth it. Olympus apparently decided it was worth it. No one else can know for sure what the results would be with or without various AA filters.
Great Bustard said:
Well, of course no "big deal" -- they're just cameras! ;) But it is interesting to know just how much more detail the weaker AA filter gives over a stronger AA filter (that's basically the bottom line in these discussions) and if that additional detail is "worth" the risk of artifacts, sometimes rather severe.
Click to expand...
I agree it is interesting, but speculative, and there is no data available to provide the certainty he seeks. Again, only Olympus knows for sure about the contribution of the AA filter versus improved processing in the E-5. No one else has the means to test that, so it is a question that can't be answered.
Great Bustard said:
Of course, that's entirely subjective. But Tim feels that if reviews are comparing identically processed E3 and E5 files, it gives a false impression as to what the real differences are for a photographer who competently processes their files.
Click to expand...
Well, no, Tim told me he sees it as technical, not subjective.

--
Bob Cole
http://www.robertcolephotography.com
 
ROC124 said:
Great Bustard said:
What Tim is asking, and what I am wondering, is that if the files were "optimally processed" for each system, rather than the same processing being applied by both, how much of a difference is really there?
Click to expand...
Only Olympus knows for sure, and I'm quite sure they won't disclose that info.
Click to expand...
Yep. Want to buy an E3 and E5, yank out their AA filters, and do a test for us? ;)
ROC124 said:
ROC124 said:
With a really sharp lens, the difference between the E3 and E5 would be 10%. However, some MTF-50 tests show a 30% differential. What Tim is questioning is the "validity" of the 30% figure.

The point he's trying to make, I think, is that the "actual" difference is a lot closer to 10% than 30%, and the risk of artifacts isn't worth the small gain he feels the weaker AA filter gives.
Click to expand...
All that is fine, but again it is subjective whether it is worth it. Olympus apparently decided it was worth it. No one else can know for sure what the results would be with or without various AA filters.
Click to expand...
I don't disagree a bit.
ROC124 said:
ROC124 said:
Well, of course no "big deal" -- they're just cameras! ;) But it is interesting to know just how much more detail the weaker AA filter gives over a stronger AA filter (that's basically the bottom line in these discussions) and if that additional detail is "worth" the risk of artifacts, sometimes rather severe.
Click to expand...
I agree it is interesting, but speculative, and there is no data available to provide the certainty he seeks. Again, only Olympus knows for sure about the contribution of the AA filter versus improved processing in the E-5. No one else has the means to test that, so it is a question that can't be answered.
Click to expand...
Again, I don't disagree. I think what Tim is seeking is not a definitive answer to a question that won't be answered, but photos of the same scene taken with the same lens on both cameras with the same settings, and "optimally processed" for both, so that the differences, whatever they are, can be given a context.

Oh wait, what was that you were saying about "a question that can't be answered"? ;)
ROC124 said:
ROC124 said:
Of course, that's entirely subjective. But Tim feels that if reviews are comparing identically processed E3 and E5 files, it gives a false impression as to what the real differences are for a photographer who competently processes their files.
Click to expand...
Well, no, Tim told me he sees it as technical, not subjective.
Click to expand...
The issue is technical; the interpretation of the results is subjective.
 
Great Bustard said:
DPR said the exact same thing. Now, even if they were typical values, given the radical differences between cameras, it does raise the question as to why there would be a typical value at all, as if sharpening doesn't require much tuning to any particular camera at all.

The entire point Tim was making is that sharpening made a marked difference between the E3 photo and E5 photo, and that he used rather different values for each.

So, combine that with your saying that DPR "tried various sharpening approaches to attempt to extract more resolution, but it didn't improve the result", then this is very curious indeed. The implication is that sharpening doesn't affect the MTF-50 score, which doesn't ring as true at all.
Click to expand...
So we can agree that DPR says they tune the USM to the camera to get the best apparent resolution. Tim (I feel odd talking about him when not present, but hope he later corrects any mischaracterization) also agreed that sharpening can't increase the measurable resolution which is the basis for the MTF-50 score), but can increase the apparent (perception of) resolution, which is his basis for perceptions of differences between the posted E-3/5 images.

It suprises me not at all that DPR has a typical USM value as a starting point. It could simply be a time saver to start with a value they may have found through experience is relatively close for the majority of cameras they test. I do exactly the same thing in LR3 and CS5 for those reasons. It shouldn't raise any questions about the methodology since they specifically state tuning the values for each camera. The only other possibility is that Tim is really questioning the competence of DPR review staff. Again that is entirely speculative and can't be answered by anyone here. We have only their reports to go by, and it appears they followed an appropriate methodology.

Of course sharpening made a marked difference in the two photos with different values for each. that is to be expected with different sensors and diferent AA filters even if the lens is the same. More importantly, one should never assume that such photos can ever adequately show the resolution differences that can be measured in lab tests, unless adequately controlled conditions and rigorous methodology are used, and such resolution differences don't show up well on many computer monitors.
--
Bob Cole
http://www.robertcolephotography.com
 
Great Bustard said:
but when did he post Oly E10's photo? Surely Oly E10 wasn't upset at Tim linking to the post with the photo, right?
Click to expand...
Peter didn't say Tim posted his image, and it does sound like he was upset at Tim's linking to it.

Well, I apologize to all for misunderstanding. It's often hard to figure out who's responding to what post here.

Julie
 
ROC124 said:
Great Bustard said:
DPR said the exact same thing. Now, even if they were typical values, given the radical differences between cameras, it does raise the question as to why there would be a typical value at all, as if sharpening doesn't require much tuning to any particular camera at all.

The entire point Tim was making is that sharpening made a marked difference between the E3 photo and E5 photo, and that he used rather different values for each.

So, combine that with your saying that DPR "tried various sharpening approaches to attempt to extract more resolution, but it didn't improve the result", then this is very curious indeed. The implication is that sharpening doesn't affect the MTF-50 score, which doesn't ring as true at all.
Click to expand...
So we can agree that DPR says they tune the USM to the camera to get the best apparent resolution.
Click to expand...
The thing is, I didn't read where they said that on the E5 or NEX7. And, now that you mention it, I went back to the E3 review and see they said the exact same thing again:

http://www.dpreview.com/reviews/olympuse3/page29.asp

So what is DPR doing?
ROC124 said:
Tim (I feel odd talking about him when not present, but hope he later corrects any mischaracterization) also agreed that sharpening can't increase the measurable resolution which is the basis for the MTF-50 score), but can increase the apparent (perception of) resolution, which is his basis for perceptions of differences between the posted E-3/5 images.
Click to expand...
What I am saying, and what I think Tim would agree with, is that the MTF-50 score is not an absolute measure of resolution but dependent upon processing, just like IMATEST's measures of DR.

If processing didn't affect the MTF-50 score, then there'd be no reason to mention the processing at all, right?
ROC124 said:
It suprises me not at all that DPR has a typical USM value as a starting point. It could simply be a time saver to start with a value they may have found through experience is relatively close for the majority of cameras they test. I do exactly the same thing in LR3 and CS5 for those reasons. It shouldn't raise any questions about the methodology since they specifically state tuning the values for each camera. The only other possibility is that Tim is really questioning the competence of DPR review staff. Again that is entirely speculative and can't be answered by anyone here. We have only their reports to go by, and it appears they followed an appropriate methodology.
Click to expand...
If DPR is using different sharpening for different systems, then why wouldn't they say exactly what settings they're using? Or are they and I missed it? But that's a bit confusing, because they have stated the processing, and it's the exact same for at least three different systems.
ROC124 said:
Of course sharpening made a marked difference in the two photos with different values for each. that is to be expected with different sensors and diferent AA filters even if the lens is the same. More importantly, one should never assume that such photos can ever adequately show the resolution differences that can be measured in lab tests, unless adequately controlled conditions and rigorous methodology are used, and such resolution differences don't show up well on many computer monitors.
Click to expand...
As I said, the MTF-50 reports lw/ph at 50% contrast. Processing (not the least of which, sharpening) affects the contrast which will necessarily affect the MTF-50 score.
 
windsprite said:
Great Bustard said:
but when did he post Oly E10's photo? Surely Oly E10 wasn't upset at Tim linking to the post with the photo, right?
Click to expand...
Peter didn't say Tim posted his image, and it does sound like he was upset at Tim's linking to it.

Well, I apologize to all for misunderstanding. It's often hard to figure out who's responding to what post here.
Click to expand...
Upset for not asking permission to link a photo? No, I don't go for that. It's one thing to embed someone's photo in a post. But linking to a photo is not only fine, it should be expected.
 
Great Bustard said:
The issue is technical; the interpretation of the results is subjective.
Click to expand...
I appreciate the distinction. However lab tests are used to ensure the results are repeatable, not subjective, and accurate. Since the desired proof here are comparitive photos on the Web, not lab tests designed to show the difference, I think it moots the technical issues.

I would be happy to shoot some rigorous non-lab photos with optimal sharpening for comparison. I've been doing that work all week with the E-3 to optimize my own prints with various lenses. My process involves a sturdy tripod, sandbag on the camera, mirror lock-up, remote release, shooting in windless conditions of a subject (local fire station) that includes variety of architectural textures and angles, and variety of vegetation including grasses, shrubs and trees some with and without leaves. Has been very instructive!

Now somebody send me an E-5 and I will get started. :)

--
Bob Cole
http://www.robertcolephotography.com
 
ROC124 said:
Great Bustard said:
The issue is technical; the interpretation of the results is subjective.
Click to expand...
I appreciate the distinction. However lab tests are used to ensure the results are repeatable, not subjective, and accurate. Since the desired proof here are comparitive photos on the Web, not lab tests designed to show the difference, I think it moots the technical issues.
Click to expand...
If the results don't represent the reason we make the measure in the first place, then there's a problem.

A good example of this is comparing two systems on the basis of default jpgs. This is a fine measure for people who shoot jpgs on default settings, but a horrible measure for people who don't.
ROC124 said:
I would be happy to shoot some rigorous non-lab photos with optimal sharpening for comparison. I've been doing that work all week with the E-3 to optimize my own prints with various lenses. My process involves a sturdy tripod, sandbag on the camera, mirror lock-up, remote release, shooting in windless conditions of a subject (local fire station) that includes variety of architectural textures and angles, and variety of vegetation including grasses, shrubs and trees some with and without leaves. Has been very instructive!

Now somebody send me an E-5 and I will get started. :)
Click to expand...
I'm pretty sure that Tim has said he will pay for an E5 for anyone who would do that for him. :D
 
Great Bustard said:
windsprite said:
Great Bustard said:
but when did he post Oly E10's photo? Surely Oly E10 wasn't upset at Tim linking to the post with the photo, right?
Click to expand...
Peter didn't say Tim posted his image, and it does sound like he was upset at Tim's linking to it.

Well, I apologize to all for misunderstanding. It's often hard to figure out who's responding to what post here.
Click to expand...
Upset for not asking permission to link a photo? No, I don't go for that.
Click to expand...
Are you saying you don't believe he said it (he did--see his post quoted below) or that you can't get behind his saying it?
Great Bustard said:
It's one thing to embed someone's photo in a post. But linking to a photo is not only fine, it should be expected.
Click to expand...
I didn't make any value judgments on either poster's words or actions. That's between Tim and Peter, as far as I'm concerned.

Julie
Oly E10 said:
A simple request to use MY image to illustrate your point would have been appropriate IMO.
No big deal really, but I guess I asked for that by not copy-writing my image.
Click to expand...
 
windsprite said:
windsprite said:
windsprite said:
windsprite said:
but when did he post Oly E10's photo? Surely Oly E10 wasn't upset at Tim linking to the post with the photo, right?
Click to expand...
Peter didn't say Tim posted his image, and it does sound like he was upset at Tim's linking to it.

Well, I apologize to all for misunderstanding. It's often hard to figure out who's responding to what post here.
Click to expand...
Upset for not asking permission to link a photo? No, I don't go for that.
Click to expand...
Are you saying you don't believe he said it (he did--see his post quoted below) or that you can't get behind his saying it?
Click to expand...
The latter. I'm amazed that someone got upset that one of his photos was linked in a technical discussion. I mean, for sure, I could see being upset if someone linked his photo and said it sucked, or something, but this? What the hell? What was there to ask permission for? It would be like me emailing PZ to ask permission to link to one of their lens tests.
windsprite said:
windsprite said:
It's one thing to embed someone's photo in a post. But linking to a photo is not only fine, it should be expected.
Click to expand...
I didn't make any value judgments on either poster's words or actions. That's between Tim and Peter, as far as I'm concerned.
Click to expand...
Absolutely it's between Tim and Peter, but I'm just amazed that it was even brought up in the first place.
 
Great Bustard said:
I'm pretty sure that Tim has said he will pay for an E5 for anyone who would do that for him. :D
Click to expand...
Yep. I will pay and pay and pay ...

Oooops, I was wrong. No I won't.

By the way, thanks for a very balanced discussion.

However, if anyone out there with two different types of cameras would post comparable images, we would really appreciate it.

By the way, ROC, for the last thread, the images I used were RAW offered to me by djbrom. I'm sure he'd share with you the same if you want to play.

And Great Bustard, you summed up my argument very well. Thanks.
 
Originally, I thought he was the owner of the image I posted. Then I realised it was Marin. Then I was confused. Now I'm clear. But, I agree with Mr Great.
 
Great Bustard said:
Are you sure it's not CA/PF?
Click to expand...
One of my friends is a professional photographer (among other things) and he has a great many cameras. One day, we set up a moire torture test among the great and the good of photographic equipment. Among them was my E3.

The E3 was the only camera in that test did not exhibit moire or other colour deviations.

This is one thing about Oly that has always endeared me: they didn't go for the easy way out if it compromised IQ. Colour is more important to me than sharpness because, for me, colour is a lot harder to correct than sharpness.

As to whether it is the result of a weak AA or something else: if one spends time staring at Kodak SLRn or SLRc files, one gets to know what alias artifacts look like because the Kodak has no AA. Because it has no AA, there are colour deviations all over the file just as they are in the E5 files I have seen. For this reason (alias artifacts, of which colour deviations is only one), no stock agency would accept Kodak files even though they are incredibly mind-blowingly sharp and, on the whole, have astounding deep rich colour.
 
rovingtim said:
Great Bustard said:
Are you sure it's not CA/PF?
Click to expand...
One of my friends is a professional photographer (among other things) and he has a great many cameras. One day, we set up a moire torture test among the great and the good of photographic equipment. Among them was my E3.

The E3 was the only camera in that test did not exhibit moire or other colour deviations.

This is one thing about Oly that has always endeared me: they didn't go for the easy way out if it compromised IQ. Colour is more important to me than sharpness because, for me, colour is a lot harder to correct than sharpness.

As to whether it is the result of a weak AA or something else: if one spends time staring at Kodak SLRn or SLRc files, one gets to know what alias artifacts look like because the Kodak has no AA. Because it has no AA, there are colour deviations all over the file just as they are in the E5 files I have seen. For this reason (alias artifacts, of which colour deviations is only one), no stock agency would accept Kodak files even though they are incredibly mind-blowingly sharp and, on the whole, have astounding deep rich colour.
Click to expand...
OK, this is interesting. Clearly, a Bayer CFA has an AA filter for a reason. Let's take a moment to walk through why a camera manufacturer would bother with an AA filter.

Each pixel is covered by either a red, blue, or green color filter. So, what happens to all the red and blue light falling on a green filter? The AA filter directs some of that light to the adjacent pixels covered with red and blue filters so it is not lost.

Of course, it's more complicated, because each color filter admits a range of colors, and these ranges overlap for each of the filters. That is, a green filter does let a certain amount of red and blue light get through. However, that red and blue light is recorded as green. Welcome to color noise!

Anyway, the processing of the RGGB pixel array takes into account (hopefully) the transmission error of the color filters and the strength (blur) of the AA filter. But, now that I think about it, how would 2nd party RAW converters have this information? Sounds like a new thread...

But the AA filter clearly serves a purpose, else the manufacturers wouldn't go to the trouble to put one in. The question is if software might not be able to do what the AA filter does, and do it better. Well, processing can't do what the AA filter does, but Olympus clearly felt that a weaker AA filter in combination with new software could.

You say that the tell-tale artifacts of an E5 file are the signature of an AA filter that is "too weak", and an unacceptable trade off for increased detail. That's a fair point that I can't deny.

All I can say is that, for the most part, the E5 files I've looked at look very good. I'm thinking that the E30 is the better system to compare against the E5. The confounding variable of pixel count is removed, and I was under the impression that the E30 did not have a weak AA filter (although, if it used the same AA filter of the 10 MP E3, that would have made it a relatively stronger AA filter on the E30 than the E3).

It's interesting to discuss, but the ultimate solution is a sensor with so many pixels that the blur of the lens itself acts as the AA filter. That way, you don't need an AA filter, but get the results of an AA filter, along with the most that you can get out of your lenses.

So, is 16 MP "enough"? Not by a long shot. Olympus should be leading the way with high MP sensors to make the best use of their lenses. I've yet to understand why they can put 14 MP on a 6x sensor (same density as 126 MP on a 4/3 sensor) but are so slow to even put 16 MP on a 4/3 sensor.

'Course, don't think I'm singling 4/3 out -- I have the same criticism for APS-C and FF.
 
You must log in or register to reply here.

Similar threads

J
JPEG chroma subsampling
Replies
1
Views
10K
xpatUSA2
xpatUSA2
T
  • User review User review
OM-5 II is a gem
Replies
0
Views
27K
Tannin
T
N
Back focus problem after E-400 disassembly
Replies
7
Views
2K
northcamel
N
bluevellet
  • User review User review
Good performance at a Bargain
Replies
1
Views
10K
DDoram
DDoram
O
Mistakes optical engineers make
Replies
12
Views
3K
Jack Hogan
J

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top