False Detail vs Real detail. MPs and AA filters and crispy lines

[bobn2]

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo

Not just 'Moire' but also other false details, look particularly in the lock of hair, where there is apparently texture running normal to the hairs. This is a very real issue. There si a fashion photographer on the Nikon from starting a thread saying he's having real and present problems with fabrics and his D810. One of the things that grates about all this is that I've been a consistent advocate of AA filters, indeed, agains some of the people who are now championing the relative lack of aliasing in the HR mode. Why the turnround.? When Olympus decided to weaken or omit AA filters to try to make up their pixel deficit, they were all for it. Now Olympus has produced a very nicely aliasing resistant mode (the E-M5II still aliases in normal mode - I suspect that this performance is simply that the lenses are giving out and acting as the AA filter) suddenly the wish to point out aliasing in other brands which have taken the same unwise route of omitting AA filters.

 

[Great Bustard]

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo

Not just 'Moire' but also other false details, look particularly in the lock of hair, where there is apparently texture running normal to the hairs. This is a very real issue. There si a fashion photographer on the Nikon from starting a thread saying he's having real and present problems with fabrics and his D810. One of the things that grates about all this is that I've been a consistent advocate of AA filters, indeed, agains some of the people who are now championing the relative lack of aliasing in the HR mode. Why the turnround.? When Olympus decided to weaken or omit AA filters to try to make up their pixel deficit, they were all for it. Now Olympus has produced a very nicely aliasing resistant mode (the E-M5II still aliases in normal mode - I suspect that this performance is simply that the lenses are giving out and acting as the AA filter) suddenly the wish to point out aliasing in other brands which have taken the same unwise route of omitting AA filters.

Obviously, I don't disagree with any of the above.

 

[alatchin]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

Indeed. This is why I said:

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo (although I did not notice any other portions of the photo that showed destructive moire to anywhere near these levels, but I allow that I may have missed it).

To my eyes, for this particular portion of the photo, the EM5II has the "most pleasing" rendering due to the absence of moire. And while the D810 crop appears to be the most detailed, a big issue here is how much of the detail is real detail and how much is fake detail. So, Bob, if you're reading, can I ask that you repost the above along with a crop from the Phase One IQ180? And, while I'm asking, could you also do the same for a few other portions of the scene? It would be most appreciated!

This exercise was to see how many can spot real detail for false detail. This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

And why Bob produced the requested crops.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length... Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

So far as I can tell, we proponents of total light, Equivalence, and any number of other facts, are being just as consistent here with regards to detail and artifacts, presenting unbiased evidence in the forms of photos, and discussing the relationship between real detail and false detail.

To be honest, I am surprised and dismayed that you, or anyone, would see it otherwise.

These conversations go around and around. I am disputing Bobs comment that comparing detail is a foolish act. Why he is the arbiter of what is foolish or not is only known to him.

However picking a camera whose level of detail is close to that of our body under question is revealing, picking a body that is miles out of the realm of reproduction for the 2 bodies under comparison works well as a correction for both.

However ina couple of threads, review etc. It seems to be repeated that the EM5.2 cannot match a 36mp sensor. However it clearly can, and it can exceed the detail reproduction accurately.

Now to suggest that all the comparison of resolution should not be compared (and to do so being foolish) to say "total light gathered" despite sensor manufacturer differences, body age etc... well that is simply a silly comment.

 

[alatchin]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

OK, which is why DPR uses the IQ180 as its reference, I suppose. But the problem is, there is no easy quantification of 'detail', which is why these discussions always get waylaid into subjective comparisons. One answer would be to do what DxOmark does, and decide that 'detail' is synonymous with 'information', for which there is a good quantitative measure.

Because the question is, can it resolve to the level of a 36mp body. Which is of interest to anyone who takes an interest in it.

This exercise was to see how many can spot real detail for false detail.

But I would argue it didn't do that, hence the arguments about what is 'false detail' and what 's 'real detail'. Is it 'real' just because it looks 'real'.? Is it 'false' just because it doesn't.

That was part of the exercise and an interesting aside that led to the poll. but more on that further down. And just because arguments start doesn't mean nothing is learned in a forum.

This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

I think that it probably does reproduce the highest level of detail of all the 36MP bodies. However, it also produces a very large amount of aliasing. In the previous threads, the comments I was taking exception to were that all the detail was false, or even that most of it was. Technically, that's simply false. What I wouldn't argue against is that the aliasing is a pain, and a bit less detail information along with a lot less aliasing, as produced by the HR mode, is in many cases (but not all) a whole load more useful.

Much of the detail is false, most of the etching lines runs in the wrong direction etc. Nevermind the moire. What impact does this have on those for whom resolution is important? Up the the individual to decide.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length...

Actually, that beat is missed quite a few times. By and large the 'correction' is made whan someone is saying something that might actively mislead, and affect equipment choices for other than the truth. Usually, when someone suggests, for instance, that a 42.5/1.2 on mFT is doing the same job as a 84/1.2 on FF. It isn't, and if someone spent the money thinking it would, they would be disappointed. If however the buy the lens knowing exactly what its performance parameters are, they'll be very happy with an excellent lens.

So it is about buyer disappointment... Well what about those who have purchased an f1.2 lens being told "aha, that is actually an F2.5 lens" omitting the idea that in such a scenario the FF body no longer has any image quality advantage. I would imagine those already woning the lens might end feeling a little hard done by. But this is irrelevant.

Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

I can't quite make out the sense of that statement. First point, what are we calling detail? If we use the DxOmark, detail=information, then it very definitely can't produce it in some senses (luminance) though it can in others (tri-channel). But in a sense that discussin is a bit pointless. It looks like the hi res mode van produce a very good image with some real advantages over a bayer 36MP. So if you want bragging rights, there they are but in the end, the quibbling about 'detail' doesn't help inform people's choices very much.

This isn't about bragging rights but fact. Either the EM5.2 can, or cannot exceed a 36mp bayer sensor. Many state it cannot, but they are considering what detail "looks like" from a high res bayer sensor. In this case it at face value seems to have all this detail absent from the EM5.2 However when you consider the nature of the high res mode and its inability to produce these artifacts it becomes very clear that in this narrow envelope it can indeed handily exceed a 36mp sensor. So the "facts" in this case are wrong.

This can be observed quite easily looking at the upressing of the 36mp image to the 63mp file size. A brief glance shows how much of the detail in that sketch is false, hence my explanation in my original post showing that the output of the EM5.2 along with explanations and an attempt to show the image differences.

Either way, I have to go home. Generally you and I end up in these things and they take up far too much time.

Have a good night.

 

[Anders W]

This is my last contribution to this 'discussion'.

In the last thread on the topic here you promised me and others that on at least three occasions without making good on your promise. So why should we trust you this time?

Well, I will break my rule this time.

Here we go again. This is the fourth time and counting.

No Anders, I didn't.

Yes you did. See below.

By 'discussion' I didn't mean the whole thread. I meant the pointless exchange with you.

It does not help to say that you didn't mean the whole thread. You continued to reply to texinwien as well as to me after having told us both that you had made your last "contribution" in your exchange with each of us.

And this really is my last contribution to this 'discussion', possibly not to this thread, or even to a different discussion with you.

No problem with that at all. Having shown that you think it is foolish to compare cameras against any metric and additionally noting that you have already managed to contradict yourself by pointing to total light as a meaningful non-arbitrary metric, e.g., here, I am more than happy to rest my case.

I live in hope that one day you might be amenable to a constructive discussion. Probably a vain hope, but there we are.

I live in the hope that one day you might realize as well as recognize what a constructive discussion actually is and discipline yourself sufficiently to take part in it. Probably a vain hope, but there we are.

 

[Anders W]

Doesn't matter how many pixel the sensor has...

Demosaicing will introduce false information because it's a little bit guess work (Bayer sensor).

I think that's part of the issue. Bayer image processing is still using algorithms designed to maximise the apparent resolution, on the basis that the sensor hasn't enough. With the newer high resolution sensors it makes more sense to aim to minimise artefacts, on the basis that the sensor has too much of those.

The problem with a less aggressive demosaic algorithm is that it will tend to reduce detail (real and false) and increase blur at the same time it tends to reduces artifacts. The trade-off is illustrated by looking at the raw and jpeg versions of the D810 studio shot. There are confounding processing differences, of course, but nonetheless such a comparison is a convenient way to get a sense of the trade-off inherent in a more aggressive demosaicing algorithm (ACR's) vs. a less aggressive one (Nikon's).

Absolutely spot-on. The point is, if you don't want the whole resolution, then why not have the less aggressive demosaicking and avoid at least some of the more objectionable artefacts (you'll never get rid of them all, which is why no-AA-filter is such a bad idea.

I like RawTherapee for those times I'm dealing with images with challenging aritifacting problems. The multiple demosaicing options gives you flexibility with how you attack the problem. I don't think there's any one ideal algorithm and the best solution will vary from image to image, although in most portions of most images you won't see any difference at all.

By the way, the objectionable artifacts issue with AA-less sensors is why I don't think it's best to try to use the D810 as the benchmark for analyzing the resolution "equivalence" of the Oly HiRes image. I think it's less confounding to use the D800 for that purpose.

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

Almost like ideas of "total light gathered"

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

I think the non-arbitrary criteria we are already using are those implicitly proposed by your OP: As much real detail as possible and as little false detail as possible. You have already provided some very good visual evidence so let me just provide some numbers as a supplement.

The criteria in question can be translated into numbers by means of MTF curves and their associated Nyquist frequencies. In the preceding thread on the subject, I posted the following such curves for the E-M5 II in hi-res mode and the D810, respectively.

Since the pixels of the E-M5 II in high-res mode have full-color information (in spite of the data being compressed into a Bayer-type format after processing the eight input shots), it is reasonable to think of a single Nyquist frequency at half a cycle (line pair) per pixel. At this point, the MTF-value (i.e., the contrast level) is practically zero which means that there will be no false detail in the form of aliasing or moiré.



Since the image data provided by the D810 are of the ordinary Bayer type, with only partial color information for each pixel, we can in this case no longer reckon with a single Nyquist frequency. Rather, as explained here, we have a range of Nyquist frequencies depending on specifically what kind of resolution we consider. For vertical and horizontal resolution in the green channel, we still have Nyquist at 0.5 cyles per pixels. For diagonal resolution in all three color channels, we have Nyquist at 0.707*0.5 = 0.3535 cycles per pixel. And for vertical and horizontal resolution in the red and blue channels, we have Nyquist at 0.5*0.5 = 0.25 cycles per pixels. As you can see from the graph below, the contrast ranges from 15 percent at the highest Nyquist frequency to no less than 40 percent at the lowest Nyquist frequency. It follows that the D810 will produce plenty of false detail whenever the subject contains detail beyond these Nyquist frequencies.



The advantage of the D810 is that it produces higher MTF values (i.e., higher microcontrast) than the E-M5 II high-res mode even for such spatial frequencies where both are capable of showing real detail. While the absolute values on the x- as well as y-axis are not directly comparable between the two graphs above (e.g., because the pixel counts are different, 64 versus 36 MP, and for reasons indicated in the technical note below), there is nevertheless no question that the D810 does better in this regard. On the other hand, as I think has been pretty convincingly demonstrated by now, that shortcoming of the E-M5 II hi-res mode can largely be compensated by sharpening. This in turn means that the main difference is in the domain where the D810 goes beyond Nyquist and therefore shows false detail whereas the E-M5 II shows real detail or, if we go sufficiently far to the right on the x-axis, no detail at all.

Technical note: The MTF curves presented above were obtained by means of MTF_Mapper using this and this RAW file from Imaging Resource. Measurements were taken from the two parallelograms closest to the center of the test chart. The curves show the average of the horizontal/sagittal and vertical/meridional MTF-values for the green channel (in order to eliminate the effect of any lateral CA) and are offered here only for the purpose of providing an approximate idea of what the curves are like relative to the associated Nyquist frequencies. The curves are not really appropriate for an evaluation of the max MTF performance of which the two systems are capable, e.g., because the E-M5 II was shot at f/8, which is clearly beyond peak for any decent FT/MFT prime (in this case the FT 50/2 Macro).

 

[Anders W]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

OK, which is why DPR uses the IQ180 as its reference, I suppose. But the problem is, there is no easy quantification of 'detail', which is why these discussions always get waylaid into subjective comparisons. One answer would be to do what DxOmark does, and decide that 'detail' is synonymous with 'information', for which there is a good quantitative measure.

Exactly how is this quantitative measure defined in precise mathematical terms and for what reasons do you think it qualifies as "good"?

This exercise was to see how many can spot real detail for false detail.

But I would argue it didn't do that, hence the arguments about what is 'false detail' and what 's 'real detail'. Is it 'real' just because it looks 'real'.? Is it 'false' just because it doesn't.

Do you have a definition of "real" and "false" detail? If so, yes. If not, why?

This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

I think that it probably does reproduce the highest level of detail of all the 36MP bodies. However, it also produces a very large amount of aliasing. In the previous threads, the comments I was taking exception to were that all the detail was false, or even that most of it was. Technically, that's simply false.

More specifically, you claimed in that prior thread, for example here, that texinwien and I had contended that all the detail shown by the Nikon D810 was false. However, neither of us had actually said anything like that and you have nowhere substantiated that we did. Moreover, you kept repeating this false accusation although I had reminded you shortly before that point, specifically here, about what I had actually said at the outset of the exchange between us about this point:

"The D810 is sharper (for luminance detail, not color) before sharpening all right. But once both are sharpened to a degree I personally find appropriate, there isn't any difference to speak of with regard to detail below a certain spatial frequency threshold, as I hope you can see from the first of the two posts I linked to [i.e. this]. For detail above that threshold, the E-M5 II hi-res shot shows less and the D810 more, but that detail is largely false. This, combined with the better rendering of color detail, makes the E-M5 II hi-res shot clearly preferable in my opinion."

In other words, what you provide here is just another example of one of your favorite fallacies: the straw man. Rather than sticking to the rules of rational discourse by taking issue with the real position of your opponents, you simply claim, without evidence, that they have said something silly and then prove that figment of your imagination wrong.

 

[Great Bustard]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

Indeed. This is why I said:

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo (although I did not notice any other portions of the photo that showed destructive moire to anywhere near these levels, but I allow that I may have missed it).

To my eyes, for this particular portion of the photo, the EM5II has the "most pleasing" rendering due to the absence of moire. And while the D810 crop appears to be the most detailed, a big issue here is how much of the detail is real detail and how much is fake detail. So, Bob, if you're reading, can I ask that you repost the above along with a crop from the Phase One IQ180? And, while I'm asking, could you also do the same for a few other portions of the scene? It would be most appreciated!

This exercise was to see how many can spot real detail for false detail. This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

And why Bob produced the requested crops.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length... Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

So far as I can tell, we proponents of total light, Equivalence, and any number of other facts, are being just as consistent here with regards to detail and artifacts, presenting unbiased evidence in the forms of photos, and discussing the relationship between real detail and false detail.

To be honest, I am surprised and dismayed that you, or anyone, would see it otherwise.

These conversations go around and around. I am disputing Bobs comment that comparing detail is a foolish act. Why he is the arbiter of what is foolish or not is only known to him.

Let's stop right there. Where did bob say or imply "comparing detail is a foolish act"? I can't imagine him saying such a thing. What Bob said was:

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

which is all together different than saying that comparing detail is a foolish act.

 

[Great Bustard]

This isn't about bragging rights but fact. Either the EM5.2 can, or cannot exceed a 36mp bayer sensor. Many state it cannot, but they are considering what detail "looks like" from a high res bayer sensor.

Given that the eight exposures effectively have 64 MP and gather twice the light as a single exposure on FF, I'm not sure why anyone would expect anything other than the EM5II hi-res mode resolving better.

However, the resolution increase does not look as large as the above suggests, and the reason is that the bulk of that has gone not into increasing luminance resolution but increasing color resolution and reducing artifacts which, I'm sure, Bob would argue is a good thing.

 

[Anders W]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

Indeed. This is why I said:

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo (although I did not notice any other portions of the photo that showed destructive moire to anywhere near these levels, but I allow that I may have missed it).

To my eyes, for this particular portion of the photo, the EM5II has the "most pleasing" rendering due to the absence of moire. And while the D810 crop appears to be the most detailed, a big issue here is how much of the detail is real detail and how much is fake detail. So, Bob, if you're reading, can I ask that you repost the above along with a crop from the Phase One IQ180? And, while I'm asking, could you also do the same for a few other portions of the scene? It would be most appreciated!

This exercise was to see how many can spot real detail for false detail. This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

And why Bob produced the requested crops.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length... Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

So far as I can tell, we proponents of total light, Equivalence, and any number of other facts, are being just as consistent here with regards to detail and artifacts, presenting unbiased evidence in the forms of photos, and discussing the relationship between real detail and false detail.

To be honest, I am surprised and dismayed that you, or anyone, would see it otherwise.

These conversations go around and around. I am disputing Bobs comment that comparing detail is a foolish act. Why he is the arbiter of what is foolish or not is only known to him.

Let's stop right there. Where did bob say or imply "comparing detail is a foolish act"? I can't imagine him saying such a thing. What Bob said was:

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

which is all together different than saying that comparing detail is a foolish act.

I asked Bob here what non-arbitrary metric he would suggest as an alternative to the one he dismissed. He replied here that:

"There doesn't have to be an alternative and neither did I suggest that there did."

In view of that reply, what grounds do you have for saying that he is not suggesting that comparing detail is a foolish act.

 

[Sergey_Green]

..

I asked Bob here what non-arbitrary metric he would suggest as an alternative to the one he dismissed. He replied here that:

"There doesn't have to be an alternative and neither did I suggest that there did."

In view of that reply, what grounds do you have for saying that he is not suggesting that comparing detail is a foolish act.

Please, go away!

 

[Sergey_Green]

..

This isn't about bragging rights but fact. Either the EM5.2 can, or cannot exceed a 36mp bayer sensor. Many state it cannot, but they are considering what detail "looks like" from a high res bayer sensor. In this case it at face value seems to have all this detail absent from the EM5.2 However when you consider the nature of the high res mode and its inability to produce these artifacts it becomes very clear that in this narrow envelope it can indeed handily exceed a 36mp sensor. So the "facts" in this case are wrong.

This can be observed quite easily looking at the upressing of the 36mp image to the 63mp file size. A brief glance shows how much of the detail in that sketch is false, hence my explanation in my original post showing that the output of the EM5.2 along with explanations and an attempt to show the image differences.

As Bob's quote goes, With any camera, the real question is what can you do with it .. So what does the fact prove to you, why is it important?

Either way, I have to go home. Generally you and I end up in these things and they take up far too much time.

See you later.

 

[Ollie 2]

Another thread where the need to be seen as "right" trumps common politeness.

Or even a shred of dignity.

 

[Anders W]

..

I asked Bob here what non-arbitrary metric he would suggest as an alternative to the one he dismissed. He replied here that:

"There doesn't have to be an alternative and neither did I suggest that there did."

In view of that reply, what grounds do you have for saying that he is not suggesting that comparing detail is a foolish act.

Please, go away!

Why do you think I should? As everyone can see, I just related what had already been said and asked GB a question that is relevant on that basis. What do you find wrong about that?

 

[alatchin]

..

This isn't about bragging rights but fact. Either the EM5.2 can, or cannot exceed a 36mp bayer sensor. Many state it cannot, but they are considering what detail "looks like" from a high res bayer sensor. In this case it at face value seems to have all this detail absent from the EM5.2 However when you consider the nature of the high res mode and its inability to produce these artifacts it becomes very clear that in this narrow envelope it can indeed handily exceed a 36mp sensor. So the "facts" in this case are wrong.

This can be observed quite easily looking at the upressing of the 36mp image to the 63mp file size. A brief glance shows how much of the detail in that sketch is false, hence my explanation in my original post showing that the output of the EM5.2 along with explanations and an attempt to show the image differences.

As Bob's quote goes, With any camera, the real question is what can you do with it .. So what does the fact prove to you, why is it important?

Either way, I have to go home. Generally you and I end up in these things and they take up far too much time.

See you later.

 

[alatchin]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

Indeed. This is why I said:

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo (although I did not notice any other portions of the photo that showed destructive moire to anywhere near these levels, but I allow that I may have missed it).

To my eyes, for this particular portion of the photo, the EM5II has the "most pleasing" rendering due to the absence of moire. And while the D810 crop appears to be the most detailed, a big issue here is how much of the detail is real detail and how much is fake detail. So, Bob, if you're reading, can I ask that you repost the above along with a crop from the Phase One IQ180? And, while I'm asking, could you also do the same for a few other portions of the scene? It would be most appreciated!

This exercise was to see how many can spot real detail for false detail. This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

And why Bob produced the requested crops.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length... Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

So far as I can tell, we proponents of total light, Equivalence, and any number of other facts, are being just as consistent here with regards to detail and artifacts, presenting unbiased evidence in the forms of photos, and discussing the relationship between real detail and false detail.

To be honest, I am surprised and dismayed that you, or anyone, would see it otherwise.

These conversations go around and around. I am disputing Bobs comment that comparing detail is a foolish act. Why he is the arbiter of what is foolish or not is only known to him.

Let's stop right there. Where did bob say or imply "comparing detail is a foolish act"? I can't imagine him saying such a thing. What Bob said was:

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

which is all together different than saying that comparing detail is a foolish act.

This is semantics. A fools game is a game which fools play, someone who is naïve. It is not world apart... In fact a game which fools play might be foolish to join

I am not getting involved in this type of conversation, that is for sure a fools game.
--
“You don’t take a photograph, you make it.” -Ansel Adams
blog.alatchinphotography(dot)com

 

[Great Bustard]

Almost, but not quite. 'Total light gathered' has a real photographic consequence. To find out more, read Richard Butler's latest excellent article

Convenient how total light gathered is not an "arbitrary metric" even though its main impact is on noise... while resolution, which can be exchanged for noise removal, or image artifacts vs detail etc... this is arbitrary... Sounds a bit like cherry picking to me and calling people fools for not seeing things your way.

Firstly, I would argue that total light gathered is not 'arbitrary' in the same sense to as was the one I was arguing against, which wasn't 'resolution' but 'how it compares with a D800'. Total light is an excellent predictor of noise in the image, and of course noise can be ameliorated one way or another, but that doesn't mean it wasn't there.

And where have I called people fools for not seeing things my way - I've called the abstract activity of trying to determine which camera is 'best' by arbitrary metrics (or indeed by non-arbitrary ones) foolish.

However pitting items against "benchmarks" is a good way of understanding to what degree detail is present.

Indeed. This is why I said:

I think the discussion is far from over. The D810 and D800 both suffer from moire in this portion of the photo (although I did not notice any other portions of the photo that showed destructive moire to anywhere near these levels, but I allow that I may have missed it).

To my eyes, for this particular portion of the photo, the EM5II has the "most pleasing" rendering due to the absence of moire. And while the D810 crop appears to be the most detailed, a big issue here is how much of the detail is real detail and how much is fake detail. So, Bob, if you're reading, can I ask that you repost the above along with a crop from the Phase One IQ180? And, while I'm asking, could you also do the same for a few other portions of the scene? It would be most appreciated!

This exercise was to see how many can spot real detail for false detail. This camera was picked as it supposedly can reproduce the highest level of detail from the batch of 36mp bodies available. A lot can be learned from an exercise such as this.

And why Bob produced the requested crops.

that's not a case of not seeing things my way, it's a case of continuing with a pointless activity for no reason other than self aggrandisement via one's equipment choices.

Which is how you see it. While the proponents of "total gathered light" dont miss a beat to correct individuals on these forums, and do it at length... Correcting the idea that the High re mode cannot produce detail to a 36mp FF sensor is foolish because it is based on resolving detail and not noise?

So far as I can tell, we proponents of total light, Equivalence, and any number of other facts, are being just as consistent here with regards to detail and artifacts, presenting unbiased evidence in the forms of photos, and discussing the relationship between real detail and false detail.

To be honest, I am surprised and dismayed that you, or anyone, would see it otherwise.

These conversations go around and around. I am disputing Bobs comment that comparing detail is a foolish act. Why he is the arbiter of what is foolish or not is only known to him.

Let's stop right there. Where did bob say or imply "comparing detail is a foolish act"? I can't imagine him saying such a thing. What Bob said was:

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

which is all together different than saying that comparing detail is a foolish act.

I asked Bob here what non-arbitrary metric he would suggest as an alternative to the one he dismissed. He replied here that:

"There doesn't have to be an alternative and neither did I suggest that there did."

In view of that reply, what grounds do you have for saying that he is not suggesting that comparing detail is a foolish act.

I don't know what Bob meant by that, but, like I said, I can't imagine Bob saying or implying that comparing detail is a foolish act. Perhaps he meant comparing detail in a qualitative manner, instead of a quantitative manner was the problem. For example, let's say System A had higher color resolution whereas System B had higher luminance resolution. How would one compare the detail between the two systems?

However, I leave this for Bob, since I don't have any answers on how to compare detail between systems that have a different balance of detail and artifacts. I'm thinking it would be similar to comparing the differences in detail from different RAW converters using different demosaicing algorithms.

Noise has a similar issue when comparing overall noise when there are large differences between luminance noise and chroma noise. For example, if System A has lower chroma noise but higher luminance noise and System B is exactly opposite, which system is more noisy?

So, my best guess is that Bob was talking about something like this, except with respect to artifacts and detail. However, as I said, I will leave it to Bob to speak for himself.

 

[alatchin]

Another thread where the need to be seen as "right" trumps common politeness.

Or even a shred of dignity.

A lot can be learned when people argue a point... And this has hardly not been civil in general.
--
“You don’t take a photograph, you make it.” -Ansel Adams
blog.alatchinphotography(dot)com

 

[knickerhawk]

Let's stop right there. Where did bob say or imply "comparing detail is a foolish act"? I can't imagine him saying such a thing. What Bob said was:

I think the idea of 'resolution equivalence' is a bit of a non starter. With any camera, the real question is what can you do with it, and trying to prove that it's better (or worse) than some other camera against an arbitrary metric is a fool's game. Unfortunately, it seems to be a fools game that some round here are inordinately fond of.

which is all together different than saying that comparing detail is a foolish act.

I asked Bob here what non-arbitrary metric he would suggest as an alternative to the one he dismissed. He replied here that:

"There doesn't have to be an alternative and neither did I suggest that there did."

In view of that reply, what grounds do you have for saying that he is not suggesting that comparing detail is a foolish act.

I don't know what Bob meant by that, but, like I said, I can't imagine Bob saying or implying that comparing detail is a foolish act. Perhaps he meant comparing detail in a qualitative manner, instead of a quantitative manner was the problem.

Except that he engaged in qualitative comparison himself in the prior thread when he disagreed with resolution claims he called "hyperbolic" and posted some crops as supporting evidence. His "fools game" comment came in response to my suggestion that the better 36mp benchmark for comparing the Oly HiRes studio image to was the D800 version instead of the D810 version. Rather than address my suggestion straight on, he latched onto the phrase I used ("resolution equivalence") and made his "arbitrary metric" comment. He never actually addressed the suggestion itself.

For example, let's say System A had higher color resolution whereas System B had higher luminance resolution. How would one compare the detail between the two systems?

Precisely why I suggested the D800 as a better benchmark than the D810.

However, I leave this for Bob, since I don't have any answers on how to compare detail between systems that have a different balance of detail and artifacts. I'm thinking it would be similar to comparing the differences in detail from different RAW converters using different demosaicing algorithms.

Again, it's less of a difference to be concerned about when the D800 is used as the benchmark instead of the D810 (or D800E).

Noise has a similar issue when comparing overall noise when there are large differences between luminance noise and chroma noise. For example, if System A has lower chroma noise but higher luminance noise and System B is exactly opposite, which system is more noisy?

That's either a disguised definitional question or simply a question of personal preference.

So, my best guess is that Bob was talking about something like this, except with respect to artifacts and detail. However, as I said, I will leave it to Bob to speak for himself.

 

[Sergey_Green]

As Bob's quote goes, With any camera, the real question is what can you do with it .. So what does the fact prove to you, why is it important?

Simple enough question with a simple enough answer. I do product work as the bulk of my professional shooting. Having the option of a high res mode and knowing exactly what it delivers is important.

And if it did not resolve up to 36MP, what would change? After all D8xx models have been around for some time now.

So despite all the limitations, this technology is a great fit for me, and for others who came find uses for the technology.

With limitations it has, is it the only way to get a hi-res image?

Beyond that as a system integrating a high res option simply increases the usefulness of the lenses I already own.

So it changes nothing then.