Pages: [1] 2   Go Down

Author Topic: Everything matters, science, etc.  (Read 8771 times)

bdosserman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 120
Everything matters, science, etc.
« on: January 27, 2012, 12:17:36 am »

First a brief introduction: I am a regular reader of LuLa, but not of the forums, so please forgive me if I'm saying anything stupidly obvious or that has already been hashed over a lot. I claim no photographic expertise, but I have recently been strongly considering upgrading to medium format, so read the "Everything Matters" essay with interest.

I have two quick comments about the essay. First is that I was reading along happily (I feel I can relate to his comments on "hyperreality", while I agree with others that it's only one possible style of photography) until I got to the image comparison, and was shocked at how bad the medium format image looked. I tried looking at it on two different computers and couldn't figure out what was going on, which drove me over here to the forums, where I eventually discovered that the problem was that my browser (Chrome) wasn't rendering the colorspace properly. If I would change one thing about the essay, I would have the jpeg redone in a colorspace which doesn't have this issue, or put a huge disclaimer that if it looks wrong, you should change browsers (actually, such a disclaimer might be helpful more generally -- I went back and subsequently looked at other images elsewhere on the site in different browsers and found they were also suffering from the issue, and I had never known it!)

Secondly, I was disturbed to find that the "smaller format camera" was apparently an iPhone. I do find the wording describing the comparison to be deeply misleading, if not outright false (as per the discussion of RAW), and would urge that the wording be clarified in the name of transparency. I don't see how being coy about the iPhone adds anything to the article other than the feeling that the author is trying to pull one over on his readers.

Anyhow, the main issue I wanted to discuss is my point of view on all the discussion of scientific (or less scientific) testing of how different cameras compare under different circumstances. Even before the present article, I was struck by the contrast between MR's article "You've Got To Be Kidding" and comments he has made in other articles to the effect that medium format seems to be recognizable even on screen (eg in the P65+ field review). I'd love to see this apparent conflict hashed out (particularly before I shell out for medium format!), ideally with more of the sort of tests carried out in the "Kidding" piece. One possibility that jumps out to me from the sample images in "Kidding" is that the scenes themselves appear to have relatively low dynamic range, and relatively low fine detail, so this may explain to some extent the contrasting experience.

Two key questions I'm curious about are: to what extent is higher resolution actually visible on smallish prints, and what is the actual comparison between dynamic ranges? These could both presumably be tested by taking good test images (i.e., with very highly detailed scenes, and large dynamic ranges) with good MFDBs and high-end DSLRs, processing and printing with normal workflow (supervised by people on "both sides", if one really wanted to be fastidious), and then having subjects (chosen to have good credentials, as in the "Kidding" article) look at the prints without being told what they're supposed to be looking for. I'd be extremely interested to read about the results of such tests (I understand that tests of this general sort have been carried out on the resolution question, and indicate that resolutions well above 300dpi give perceptible differences, but it seems there's still some controversy about this?).

I am interested to hear about the scientific/engineering reasons why something should or should not be perceptibly different/better than something else, but I always take such arguments with a grain of salt. I am scientifically oriented by nature, but it's also easy to get carried away by coming up with particularly measurements and theories and stacking a bunch of them together to make a conclusion. Although this will have a veneer of scientific authority, in my opinion it is much more scientific to actually carry out a well-controlled experiment, as for instance described above. Emil Martin's forum piece trying to combine DxO's various measurements to extract a photographic dynamic range number looks like a first-rate example of the first approach, and far better than nothing, but in my opinion would be definitively trumped by a carefully controlled print comparison showing contrary results. I feel that there's a huge discrepancy between claims of roughly comparable DR versus a six-stop difference, and I can't imagine it would be hard to figure out which is right by testing. I am naturally skeptical, but of both approaches: I appreciate how knowing what you're looking at can bias your perception on the one hand, but on the other hand feel there's plenty of history where scientists/engineers think they've worked out all the relevant measurements, only to discover later they were missing some crucial aspect of the situation.

One last comment: in some sense, only positive results of controlled tests as above are definitive. A negative result only says that the people being asked to look couldn't tell the difference between the particular cameras used for the particular scenes used. Such tests can never prove that two sensors are always equivalent for a given print size. However, if a given test is trending towards negative, the obvious thing to do would be to run it again, but with someone who feels strongly that they see such differences (such as MD) as the viewing subject (and not, obviously, involved in preparing the prints). I've experienced firsthand that different people have differently discerning perception (probably based largely on experience and training), so it's a valid objection that a negative test might not have involved the "right people". But at the same time, from my point of view it's also worthwhile to know which differences are or are not readily apparent to the average educated viewer. So I feel like any such tests will be interesting.

OK, hope that wasn't too long-winded.

Brian
Logged

Eric Myrvaagnes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 22814
  • http://myrvaagnes.com
    • http://myrvaagnes.com
Re: Everything matters, science, etc.
« Reply #1 on: January 27, 2012, 09:36:43 am »

Very sensible, Brian.

Your comments make more sense and have more value IMHO than the entire content of the other threads that responded to Mark's piece.

This needed to be said, and you expressed it very well.

Eric
Logged
-Eric Myrvaagnes (visit my website: http://myrvaagnes.com)

bdosserman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 120
Re: Everything matters, science, etc.
« Reply #2 on: January 27, 2012, 11:10:26 am »

I'd recommend before you shell out for medium format you rent and do your own testing in a way that is relevant to your own work rather than relying on internet opinion.

Good point. I should perhaps have clarified a few points relative to my own situation:
- I definitely plan to test any systems before I buy, however I have no expertise in print making and little in photoshop, so I wouldn't trust my own judgement if I just tried to rent something and make some prints from it. Thus my interest in testing carried out by really well-qualified professionals
- I am interested in medium format primarily for the ability to make large prints, which is not very controversial; the other questions interest me, but are not the driving factor
- If the first item sounds strange in the context of considering medium format, I am thinking of it at least partially in the context of "future proofing" my work, so that once I get more expertise on the processing and printing side I will have the ability to go back and try to really get the most out of the photos I will have already taken. (There are also personal factors involved in the timing, bit I won't get into those)

Brian
Logged

welder

  • Newbie
  • *
  • Offline Offline
  • Posts: 40
Re: Everything matters, science, etc.
« Reply #3 on: January 27, 2012, 03:01:50 pm »

Secondly, I was disturbed to find that the "smaller format camera" was apparently an iPhone. I do find the wording describing the comparison to be deeply misleading, if not outright false (as per the discussion of RAW), and would urge that the wording be clarified in the name of transparency. I don't see how being coy about the iPhone adds anything to the article other than the feeling that the author is trying to pull one over on his readers.
I have the same sentiments on this point. It would seem a strange comparison since an iphone does not function as a primarily photographic device with a serious lens.

Now, I don't want to enter into go off on a tangent about being able to discern between fine wines, but I bring it up because it's a metaphor used in the article and I like metaphors. I know that I can tell the difference in quality between a $2.99 bottle of wine and a decent $20 of wine. But I doubt I could discern much difference in quality between a $20 bottle and a $200 bottle. Whether I could eventually be trained to do so is not an issue I want to debate; rather the point is that while I can't really tell the difference between very good and ultra high end, I can tell the difference between very good and poor. And to extend the metaphor, comparing a MFDB to an iphone is like comparing a $200 bottle of wine to a $2.99 bottle of wine.

I have no doubt that medium format digital back provides better image quality. But, this comparison I find rather underwhelming in convincing me of substantial benefits for web resolutions. (And if someone were using a MF camera primarily to produce web images, I would indeed think of that as a waste. Or at the very least as overkill.)

As far as the question of does a MF produce significantly better results for small prints, I think a more controlled "scientific" comparison to a DSLR with a good lens would be appropriate. The results may well be a little better. But are they “far superior” as the article states? I think that’s an open question. (And given previous a previous article published on this site that discussed how favorably prints from a Canon G10 compared to MFDB, I think that’s a fair topic for debate).


Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Everything matters, science, etc.
« Reply #4 on: January 28, 2012, 05:26:45 am »

Hi,

Thanks for your posting...

I would like to add a few points. First I'd like to point out that I'm shooting DSLR only. On the other hand I have some experience with MF, MF scanning and so on and also have some scientific education (enough to know about scientific method).

For me the obvious advantage of MF is higher resolution. No question about that!

The next question is if the difference between say a 40 MP MDB and a 24 MP DSLR can be seen in small prints. I would say that it depends on a few factors.

1) If the print actually carries the information there will obviously be a difference, the question is if it can be seen. It seems that normal vision is pretty insensitive to image detail past 180 PPI, a figure frequently cited by Jeff Schewe. A 24 MP DSLR will print 33" wide at 180 PPI, so you would probably need to print larger than that to see more detail. On the other hand we could print at 720 PPI and examine the print with a loupe, in that case the DSLR would be limited to 8.3" width.

2) If the print is scaled down to smaller size information will be lost and actually replaced with a lot of artifacts. Downscaling is actually bad, and prone to production of fake detail. This is a very good reason to print at high resolution.

The arcticle here gives some information: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/62-scaling-up-or-down

I really recommend these two references by Bart van der Wolf
http://bvdwolf.home.xs4all.nl/main/foto/down_sample/down_sample.htm
http://bvdwolf.home.xs4all.nl/main/foto/down_sample/example1.htm

So in short, I presume that in many cases the differences seen in small prints are resizing artifacts.

3) As an MFDB has larger size any item would show up in larger size on the sensor. We could also say that the larger sensor needs less magnification of the image for a given print size. As optics work a larger feature will have higher MTF (fine detail contrast) as MTF drops with feature size (for small features). It is also possible that MF certain lenses are simply better made than certain DSLR lenses.

Regarding DR it is like this:

A larger sensor collects more photons, that gives a certain advantage in both SNR (signal noise ratio) and SNR. DSLR sensors have much lower readout noise than the CCDs used in MFDBs so modern DSLRs are probably a good match for MFDBs.

If we assume that exposure is fully ETTR (Expose To The Right) DR would define the amount of shadow detail we could extract. Would MFDBs really have better DR than DSLRs than they would perform better than DSLRs at high ISO which they clearly don't.

Some DSLRs have quite noisy Analog Digital Converters and achieve their high ISO by preamplification of sensor signal. These systems have comparably poor DR at low ISO. All Canon DSLRs have this characteristic, also shared by Nikon D3s and D700.

Check out this link for an interesting comparison: http://www.luminous-landscape.com/forum/index.php?topic=50895.0

Best regards
Erik


First a brief introduction: I am a regular reader of LuLa, but not of the forums, so please forgive me if I'm saying anything stupidly obvious or that has already been hashed over a lot. I claim no photographic expertise, but I have recently been strongly considering upgrading to medium format, so read the "Everything Matters" essay with interest.

I have two quick comments about the essay. First is that I was reading along happily (I feel I can relate to his comments on "hyperreality", while I agree with others that it's only one possible style of photography) until I got to the image comparison, and was shocked at how bad the medium format image looked. I tried looking at it on two different computers and couldn't figure out what was going on, which drove me over here to the forums, where I eventually discovered that the problem was that my browser (Chrome) wasn't rendering the colorspace properly. If I would change one thing about the essay, I would have the jpeg redone in a colorspace which doesn't have this issue, or put a huge disclaimer that if it looks wrong, you should change browsers (actually, such a disclaimer might be helpful more generally -- I went back and subsequently looked at other images elsewhere on the site in different browsers and found they were also suffering from the issue, and I had never known it!)

Secondly, I was disturbed to find that the "smaller format camera" was apparently an iPhone. I do find the wording describing the comparison to be deeply misleading, if not outright false (as per the discussion of RAW), and would urge that the wording be clarified in the name of transparency. I don't see how being coy about the iPhone adds anything to the article other than the feeling that the author is trying to pull one over on his readers.

Anyhow, the main issue I wanted to discuss is my point of view on all the discussion of scientific (or less scientific) testing of how different cameras compare under different circumstances. Even before the present article, I was struck by the contrast between MR's article "You've Got To Be Kidding" and comments he has made in other articles to the effect that medium format seems to be recognizable even on screen (eg in the P65+ field review). I'd love to see this apparent conflict hashed out (particularly before I shell out for medium format!), ideally with more of the sort of tests carried out in the "Kidding" piece. One possibility that jumps out to me from the sample images in "Kidding" is that the scenes themselves appear to have relatively low dynamic range, and relatively low fine detail, so this may explain to some extent the contrasting experience.

Two key questions I'm curious about are: to what extent is higher resolution actually visible on smallish prints, and what is the actual comparison between dynamic ranges? These could both presumably be tested by taking good test images (i.e., with very highly detailed scenes, and large dynamic ranges) with good MFDBs and high-end DSLRs, processing and printing with normal workflow (supervised by people on "both sides", if one really wanted to be fastidious), and then having subjects (chosen to have good credentials, as in the "Kidding" article) look at the prints without being told what they're supposed to be looking for. I'd be extremely interested to read about the results of such tests (I understand that tests of this general sort have been carried out on the resolution question, and indicate that resolutions well above 300dpi give perceptible differences, but it seems there's still some controversy about this?).

I am interested to hear about the scientific/engineering reasons why something should or should not be perceptibly different/better than something else, but I always take such arguments with a grain of salt. I am scientifically oriented by nature, but it's also easy to get carried away by coming up with particularly measurements and theories and stacking a bunch of them together to make a conclusion. Although this will have a veneer of scientific authority, in my opinion it is much more scientific to actually carry out a well-controlled experiment, as for instance described above. Emil Martin's forum piece trying to combine DxO's various measurements to extract a photographic dynamic range number looks like a first-rate example of the first approach, and far better than nothing, but in my opinion would be definitively trumped by a carefully controlled print comparison showing contrary results. I feel that there's a huge discrepancy between claims of roughly comparable DR versus a six-stop difference, and I can't imagine it would be hard to figure out which is right by testing. I am naturally skeptical, but of both approaches: I appreciate how knowing what you're looking at can bias your perception on the one hand, but on the other hand feel there's plenty of history where scientists/engineers think they've worked out all the relevant measurements, only to discover later they were missing some crucial aspect of the situation.

One last comment: in some sense, only positive results of controlled tests as above are definitive. A negative result only says that the people being asked to look couldn't tell the difference between the particular cameras used for the particular scenes used. Such tests can never prove that two sensors are always equivalent for a given print size. However, if a given test is trending towards negative, the obvious thing to do would be to run it again, but with someone who feels strongly that they see such differences (such as MD) as the viewing subject (and not, obviously, involved in preparing the prints). I've experienced firsthand that different people have differently discerning perception (probably based largely on experience and training), so it's a valid objection that a negative test might not have involved the "right people". But at the same time, from my point of view it's also worthwhile to know which differences are or are not readily apparent to the average educated viewer. So I feel like any such tests will be interesting.

OK, hope that wasn't too long-winded.

Brian
Logged
Erik Kaffehr
 

EricWHiss

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2639
    • Rolleiflex USA
Re: Everything matters, science, etc.
« Reply #5 on: January 30, 2012, 01:47:51 am »

On the top of the list of what matters, I'd add:

1) Idea/Concept
2) Composition/execution

If you don't have those two then nothing else on the list actually matters.
Logged
Rolleiflex USA

crix

  • Newbie
  • *
  • Offline Offline
  • Posts: 1
Re: Everything matters, science, etc.
« Reply #6 on: February 02, 2012, 08:35:36 am »

Dear Michael,

I really enjoy reading articles on LL because I feel they are - even when reviewing gear - more related to the fun of photography and less to the gritty world of technology.

Although others seem to have commented a lot on the Article Everything matters, I feel the need to comment myself, too.

I was frankly appalled by this article. I myself have been an High end audio listener, even constructed and successfully sold High end audio equipment. In addition, I am an electrotechnical engineer.
Yes you can build a preamp for about 200$ materials audibly better than a 1000$ commerical one. But it is entirely "esoteric" BS about those cables. And the author is allowed to rant much more in this fashion. Many peoples are NOT able to discern small differences neither in audio, wine, or photography because either they are not able to or simply do not care.

Those people are happy with their pictures because they are able to see the ESSENCE it shows and don't care about the details at all. Best example is that many professional musicians do not own a very good audio system.

What also is detrimental to the advancement of photographic knowledge is his entirely unobjective and unscientific approach. If you are willing to, it would be easy to photograph a subject with a good APS-C camera & lens in a way indistinguishable from one taken with MF hardware.

A viewer of a picture may notice low DOF that emphasises the main subject. But whether that picture was shot with MF or an "amateur" DSLR will be negligible.

This article does not fit into your otherwise interesting and sound series, e.g. on the Nex 5n or 7.
Please reconsider whether you really want to continue in this direction.
Best regards,
Chris
Logged

viewfinder

  • Full Member
  • ***
  • Offline Offline
  • Posts: 124
Re: Everything matters, science, etc.
« Reply #7 on: February 02, 2012, 12:42:56 pm »

Chris,......nicely put!   I could'nt agree more!
Logged

Isaac

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3123
Re: Everything matters, science, etc.
« Reply #8 on: February 06, 2012, 08:08:07 pm »

After so much has been said about the essay, I decided to take another look and be charitable and search for some agreement:

Quote
Everything matters.  It is all about the small details.
Let's just say Yes! Now we are overwhelmed by innumerable details - How does that help?

Quote
Do not ignore details. Even the smallest ones can be crucial. The weakest link in a chain will always determine the ultimate quality of the entire chain.
Yes, for some particular qualities relevant to some particular motivation. Let's not gloss over the details about those different qualities with a generality: ultimate quality.

Quote
Do not rely on specifications, measurements or marketing claims.
Yes, take a test-drive and remember caveat emptor!

Quote
Search for the unseen.
Yes, novelty attracts our attention.

Quote
It is not only the subject matter that contains the unseen.
Yes, tonal patterns may be seen in monochrome that may not be seen when the picture is in color.

Quote
Prints and screen images from larger format captures always look better, regardless of the size of the print or the screen image.
Let's not ignore the details - it depends on who's doing the looking, and "look(s) better" may not be the quality we care about most.


Quote
For reasons I cannot understand, the vast majority of people have a tendency to underestimate the capabilities of human senses.
Is it understandable that, for each of us, the capabilities of our senses are our ordinary - the yardstick by which we estimate others' capabilities.
« Last Edit: February 08, 2012, 12:54:21 pm by Isaac »
Logged

bdosserman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 120
Re: Everything matters, science, etc.
« Reply #9 on: February 10, 2012, 10:26:05 am »

Hi,
    Thanks for your reply.

For me the obvious advantage of MF is higher resolution. No question about that!

Yes, this drives my interest as well. Of course, it will be very interesting to compare the new Nikons to low-end MF in this regard (especially with the no-AA-filter option).

The next question is if the difference between say a 40 MP MDB and a 24 MP DSLR can be seen in small prints. I would say that it depends on a few factors.

1) If the print actually carries the information there will obviously be a difference, the question is if it can be seen. It seems that normal vision is pretty insensitive to image detail past 180 PPI, a figure frequently cited by Jeff Schewe. A 24 MP DSLR will print 33" wide at 180 PPI, so you would probably need to print larger than that to see more detail. On the other hand we could print at 720 PPI and examine the print with a loupe, in that case the DSLR would be limited to 8.3" width.

I am leery of these calculations for a number of reasons:
1) bjanes quotes Schewe in another thread as saying that the difference between 360dpi prints and 720dpi prints is readily visible, with addition backup research. Not sure why the conflict, but I could image it would be based on a number of factors, including quality/resolution of printing, and the distinction between the eye being able to resolve a certain resolution versus just perceiving that there's a difference.
2) View distance is usually treated as a constant, but for me at least, when I see a print which looks especially detailed, I have a natural urge to move closer to appreciate the detail. Viewing photos isn't like sitting in a movie theater in a fixed position.
3) The advertised resolution of most digital cameras is kind of questionable in the sense that the full data is extrapolated from the bayer matrix data (not to mention the AA filters on most cameras). The actual amount of input data is only 1/3 as much as a TIFF of the same resolution holds.
4) See below.

Again, for me I'd like to see blind comparisons of actual photos, as I'll find those more convincing than theoretical discussion.

Random related question: would lens diffraction be more of an issue for smaller or larger format sensors, given equal resolution?

2) If the print is scaled down to smaller size information will be lost and actually replaced with a lot of artifacts. Downscaling is actually bad, and prone to production of fake detail. This is a very good reason to print at high resolution.

The arcticle here gives some information: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/62-scaling-up-or-down

I really recommend these two references by Bart van der Wolf
http://bvdwolf.home.xs4all.nl/main/foto/down_sample/down_sample.htm
http://bvdwolf.home.xs4all.nl/main/foto/down_sample/example1.htm

So in short, I presume that in many cases the differences seen in small prints are resizing artifacts.

I have a few thoughts on this. The first is that my understanding is that in fact, oversampling is usually the recommended method for producing highest quality analog to digital conversion, whether in audio or video. For instance, when producing top-grade Blurays they scan at way higher resolution and res down, and when making a basic ray tracer, the standard algorithm is to sample several times per pixel (eg, break up into four subpixels, then sample once in each with some random jitter to avoid artifacting). So I don't think the potential problem is ressing down per se, but rather how it's done. To be fair, I think that ressing down by fractional ratios is probably harder, and I'd imagine this comes up regularly when printing.

Also, regarding artifacts: I'm as familiar with and wary of visual artifacts as anyone (or at least, as anyone who doesn't deal with them on a professional basis), but let me step back for a minute and try to rethink the orthodoxy for a minute. I think perhaps it's fair to say that our notion of artifact is somewhat subjective, and the accepted "right" answer for what is good and bad may not be as objective as we make it out to be, especially when we're talking about detail at the very limits of human perception. Let's look at the example you posted. There's this red and white awning that's a big issue for downsizing. The accepted right way to downsize to resolutions which are incapable of rendering individual bars of color is to go for something like a uniform pink color. However, I think one could argue that a downsizing which preserved red and white pixels (even with artificial patterns, as would almost certainly occur) is in some sense as faithful, or differently faithful, to the detail in the individual image. After all, there is no actual pink in the image. So we are misrepresenting the patterns of color (which the pink version does too, but less acutely), but more accurately representing which colors are actually present. I could imagine this phenomenon playing into the perceived higher detail of MF prints (and what MR referred to as "microcontrast"). For me, ultimately, if "artifacts" are playing into what makes these prints look better, that doesn't automatically invalidate that they are perceived as looking better, especially if the artifacts are in some sense ghosts of actual detail that was captured.

Some DSLRs have quite noisy Analog Digital Converters and achieve their high ISO by preamplification of sensor signal. These systems have comparably poor DR at low ISO. All Canon DSLRs have this characteristic, also shared by Nikon D3s and D700.

Do you have any idea whether the D800 will have this issue as well? Which DSLRs don't have it?

Best,

Brian
Logged

dreed

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1715
Re: Everything matters, science, etc.
« Reply #10 on: February 11, 2012, 02:02:24 pm »

Now, I don't want to enter into go off on a tangent about being able to discern between fine wines, but I bring it up because it's a metaphor used in the article and I like metaphors. I know that I can tell the difference in quality between a $2.99 bottle of wine and a decent $20 of wine. But I doubt I could discern much difference in quality between a $20 bottle and a $200 bottle. Whether I could eventually be trained to do so is not an issue I want to debate; rather the point is that while I can't really tell the difference between very good and ultra high end, I can tell the difference between very good and poor. And to extend the metaphor, comparing a MFDB to an iphone is like comparing a $200 bottle of wine to a $2.99 bottle of wine.

Depending on the yields, etc, "expensive wine" (or grapes/juice) can often end up in boxed/casked wine and had very cheaply (although it doesn't get marked as such.)

The truth is that whether wine A or B is better depends on the person. Just like some people prefer dark chocolate to milk chocolate and others milk to dark, so too there is often a similar difference in the perceived relative merits of a wine. Now obviously there are some cases where there is a clear preference for A over B when it comes to cheap vs more expensive. So not only may the nuances of the $200 bottle be lost on someone but they might actually prefer the taste of a cheaper wine. Yes, that's heretical, I know, but after 5 years of regular tastings, that was one of the more interesting take-aways.

Quote
I have no doubt that medium format digital back provides better image quality. But, this comparison I find rather underwhelming in convincing me of substantial benefits for web resolutions. (And if someone were using a MF camera primarily to produce web images, I would indeed think of that as a waste. Or at the very least as overkill.)

As far as the question of does a MF produce significantly better results for small prints, I think a more controlled "scientific" comparison to a DSLR with a good lens would be appropriate. The results may well be a little better. But are they “far superior” as the article states? I think that’s an open question. (And given previous a previous article published on this site that discussed how favorably prints from a Canon G10 compared to MFDB, I think that’s a fair topic for debate).

The impression I get is that people are struggling for ways in which to describe how or explain why a MFDB produces a better image than an iPhone or even a DSLR.

I've seen the argument about "bigger pixels" used. The IQ180 has smaller pixels than a 5D Mark II but yet delivers better results (according to DxO.) Where's the discussion of this? I mean there has got to be a reason for this, doesn't there? Or does explaining this and accepting that smaller pixels can produce better results than bigger ones threaten too many digital photography myths?
Logged

PierreVandevenne

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 512
    • http://www.datarescue.com/life
Re: Everything matters, science, etc.
« Reply #11 on: February 11, 2012, 06:07:50 pm »

There's one overriding factor and that's read noise, as discussed in many threads already. Tremendous benefits have been achieved with the reduction of read noise from 16e to 2-3e. The gains are obvious from the engneering S/N equation, which can eventually be matched by the real world performance of a camera but never exceeded (at least with real data, eye pleasing but inaccurate data has its value in photography) FWC, the other factor, is essentially linked to pixel size. Micro lens design and process improvements have also played a part, but that's harder to quantify since the data is not as readily available or measurable. But small pixels aren't necessarily bad up to a point, as explained by Emil, because what matters is the total amount of signal collected and larger sensors will capture more signal at the price of larger lenses.

So yes, all other things being equal MF will "beat" a smaller sensor camera. A smaller sensor camera may catch up if it has a newer sensor, but will again be "beaten" by the larger sensor camera once it is upgraded.

The reason why there's a debate now between top of the line DSLRs (at this point, I am not including Canon in that category from a sensor point of view, even if I am a Canon owner) and MF is that sensors used in MF are a bit behind the current "sonpenkon" sensors. What they gain in signal collection area, they lose it - and possibly more - in read noise.

At some point in the future, as limits are reached on smaller sensors, MF sensors will catch up and we'll be able to say "same pixels, more signal collection area = better" for good.
Logged

dreed

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1715
Re: Everything matters, science, etc.
« Reply #12 on: February 11, 2012, 08:35:15 pm »

At some point in the future, as limits are reached on smaller sensors, MF sensors will catch up and we'll be able to say "same pixels, more signal collection area = better" for good.

Indeed! And to some extent, the lens plays a large part in this too.

The small lens for a mobile phone or compact digital camera simply cannot gather as much light as larger format lenses can.

It is worth going back to read this:
http://www.luminous-landscape.com/essays/dxomark_sensor_for_benchmarking_cameras.shtml
... it would be a great if the graphs in that essay were updated somewhere, from time to time, to add in newer cameras (and/or exclude older ones.)
Logged

Pete Ferling

  • Guest
Re: Everything matters, science, etc.
« Reply #13 on: February 12, 2012, 12:56:15 am »

+1 for the lens.  It's half the equation and seldom mentioned in all the those "discussions" that drives forum admins to close threads and drink. :)

I can only discuss my points based on actual use, for me I still shoot with a relic of a camera, a Mamiya M645 and use Portra 160VC exclusively.  While I have shots that were better on a crop sensor, much has to do with way too many variations that affect the outcome.  My primary use for the MF is to avoid the hassle of pan and stitch and bracketing with crop or small sensors.  I always keep the MF handy in case I should happen upon a scene that warrants it, (It's much easier to take one shot vs. nine).  The other point is a personal liking for old school trust of ones settings, and finally I just like the look of the combination. 

Its not my business to convince another which is best, and I applaud anyone for simply finding out for themselves by actually trying it. In the end, does all that science and fact work out for you?
Logged

welder

  • Newbie
  • *
  • Offline Offline
  • Posts: 40
Re: Everything matters, science, etc.
« Reply #14 on: February 14, 2012, 12:25:02 pm »

The impression I get is that people are struggling for ways in which to describe how or explain why a MFDB produces a better image than an iPhone or even a DSLR.

I've seen the argument about "bigger pixels" used. The IQ180 has smaller pixels than a 5D Mark II but yet delivers better results (according to DxO.) Where's the discussion of this? I mean there has got to be a reason for this, doesn't there? Or does explaining this and accepting that smaller pixels can produce better results than bigger ones threaten too many digital photography myths?
I actually don't care that much about describing "how or why" a MFDB image is better. I am more interested in evaluating the subjective margin by which it is judged to be better. Where is there a significant difference evident in real world viewing? Saying a MFDB will provide better results for any size print or screen image is fine, but I am left unconvinced by the article that this improvement is always significant or discernable (at least if you were to compare a MFDB to a real camera with a decent lens, and not an iphone).
Logged

Rob C

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 24074
Re: Everything matters, science, etc.
« Reply #15 on: February 14, 2012, 02:07:41 pm »

+1 for the lens.  It's half the equation and seldom mentioned in all the those "discussions" that drives forum admins to close threads and drink. :)

I can only discuss my points based on actual use, for me I still shoot with a relic of a camera, a Mamiya M645 and use Portra 160VC exclusively.  While I have shots that were better on a crop sensor, much has to do with way too many variations that affect the outcome.  My primary use for the MF is to avoid the hassle of pan and stitch and bracketing with crop or small sensors.  I always keep the MF handy in case I should happen upon a scene that warrants it, (It's much easier to take one shot vs. nine).  The other point is a personal liking for old school trust of ones settings, and finally I just like the look of the combination. 

Its not my business to convince another which is best, and I applaud anyone for simply finding out for themselves by actually trying it. In the end, does all that science and fact work out for you?


Two excedlent points, Peter, and more or less the reasons why I still live with non-af optics where I can, and why I keep my machines in as manual a configuration as I can find!

Rob C

dreed

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1715
Re: Everything matters, science, etc.
« Reply #16 on: February 14, 2012, 03:50:10 pm »

I actually don't care that much about describing "how or why" a MFDB image is better. I am more interested in evaluating the subjective margin by which it is judged to be better. Where is there a significant difference evident in real world viewing? Saying a MFDB will provide better results for any size print or screen image is fine, but I am left unconvinced by the article that this improvement is always significant or discernable (at least if you were to compare a MFDB to a real camera with a decent lens, and not an iphone).

There's also the question of whether or not the improvement is desirable..

Every time I read that someone needs a high end camera or lens to produce something worthy of art, I'm reminded of the review Michael once did of a throwaway digital camera with which he took a photo of the rainy conditions outside of his car and to which his wife remarked that it was the best piece of art he'd produced in a long time. Unfortunately I cannot currently find the link...
Logged

welder

  • Newbie
  • *
  • Offline Offline
  • Posts: 40
Re: Everything matters, science, etc.
« Reply #17 on: February 14, 2012, 05:39:54 pm »

There's also the question of whether or not the improvement is desirable..

Every time I read that someone needs a high end camera or lens to produce something worthy of art, I'm reminded of the review Michael once did of a throwaway digital camera with which he took a photo of the rainy conditions outside of his car and to which his wife remarked that it was the best piece of art he'd produced in a long time. Unfortunately I cannot currently find the link...
From personal experience I agree without hesitation that a high end gear is not required to make art. First photo I ever sold in a gallery came from a point and shoot camera. To go back to an audio metaphor: regardless of how good my speakers are, what really matters is the music coming out of them, not the speakers themselves. A MFDB may provide more details in certain viewing scenarios, but without a good composition or thoughtful artistic intent the photo is not worth anything.

Although of course for me a better quality camera is always desirable, but it's a question as to whether or not it is practical compared to what I'm using already for the work I'm currently doing. Maybe it's because I don't have a connoisseur's mindset but I think a MFDB would only enhance my photos at the margins, rather than making a substaintial impact. So at this stage in both my commerical and personal endeavors it's not an investment worth making. Although once I'm wealthy enough that may change :)

Logged

hjulenissen

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2051
Re: Everything matters, science, etc.
« Reply #18 on: February 16, 2012, 04:25:33 am »

I am leery of these calculations for a number of reasons:
1) bjanes quotes Schewe in another thread as saying that the difference between 360dpi prints and 720dpi prints is readily visible, with addition backup research. Not sure why the conflict, but I could image it would be based on a number of factors, including quality/resolution of printing, and the distinction between the eye being able to resolve a certain resolution versus just perceiving that there's a difference.
I am certain that there is more to print sharpness/detail than the number of dpi. If you really want to establish good lower bounds based on empiry, you probably dont want to use a general purpose image processing pipeline that includes god knows how many image scalings, an unknown dithering pattern, and unknown ink splatter characteristics. Not that scaling in a non-linear color/intensity representation (such as sRGB) actually gives some (usually small) errors compared to moving the image closer/further away from the viewer.

I think that image perception research have suggested that 1 minute of arc is a good rule of thumb for a normal-good pair of eyes. I presume this is for relatively high-contrast black/white transitions perfectly smoothly/sharp printed on paper. This could be compared to actual prints at a certain size/distance/resolution, and if they are in the same ballpark, I would be satisfied.
Quote
2) View distance is usually treated as a constant, but for me at least, when I see a print which looks especially detailed, I have a natural urge to move closer to appreciate the detail. Viewing photos isn't like sitting in a movie theater in a fixed position.
Sure. But when do you stop? Is it worthwhile to have an image that prints 1x1 meter, where you can step closer until you reach the close-focus limit of your eyesight and still see no resolution limitations?
Quote
3) The advertised resolution of most digital cameras is kind of questionable in the sense that the full data is extrapolated from the bayer matrix data (not to mention the AA filters on most cameras). The actual amount of input data is only 1/3 as much as a TIFF of the same resolution holds.
"Megapixels" is a horrible concept. But I do believe that Bayer sensors behave very good for the kind of scenes that photographers tend to shoot, and for viewers that are human. Certainly a lot better than what a simplistic analysis of the red/green/blur sensels would suggest.
Quote
Again, for me I'd like to see blind comparisons of actual photos, as I'll find those more convincing than theoretical discussion.
Does it have to be either or? I think that doing both and relating them to each other is a good approach.
Quote
I have a few thoughts on this. The first is that my understanding is that in fact, oversampling is usually the recommended method for producing highest quality analog to digital conversion, whether in audio or video.
Oversampling is a method where you can replace a complex analog filter with a (equally complex but simpler to design) digital filter. Or you might see it (with dithering) as a method to provide many levels at lower sample-rates, when the physical technology is better suited for fewer levels and higher sample-rates.
Quote
For instance, when producing top-grade Blurays they scan at way higher resolution and res down, and when making a basic ray tracer, the standard algorithm is to sample several times per pixel (eg, break up into four subpixels, then sample once in each with some random jitter to avoid artifacting). So I don't think the potential problem is ressing down per se, but rather how it's done. To be fair, I think that ressing down by fractional ratios is probably harder, and I'd imagine this comes up regularly when printing.
Scaling your typical over-sampled image is not that hard. Scaling an image that has been excessively over-sharpened/aliased through non-Nyquistian sampling is harder.

Many image processing algorithms and printer drivers seems to use very naiive image scaling algorithms. Perhaps because most of their customers dont care that much.
Quote
Also, regarding artifacts: I'm as familiar with and wary of visual artifacts as anyone (or at least, as anyone who doesn't deal with them on a professional basis), but let me step back for a minute and try to rethink the orthodoxy for a minute. I think perhaps it's fair to say that our notion of artifact is somewhat subjective, and the accepted "right" answer for what is good and bad may not be as objective as we make it out to be, especially when we're talking about detail at the very limits of human perception. Let's look at the example you posted. There's this red and white awning that's a big issue for downsizing. The accepted right way to downsize to resolutions which are incapable of rendering individual bars of color is to go for something like a uniform pink color. However, I think one could argue that a downsizing which preserved red and white pixels (even with artificial patterns, as would almost certainly occur) is in some sense as faithful, or differently faithful, to the detail in the individual image. After all, there is no actual pink in the image. So we are misrepresenting the patterns of color (which the pink version does too, but less acutely), but more accurately representing which colors are actually present. I could imagine this phenomenon playing into the perceived higher detail of MF prints (and what MR referred to as "microcontrast"). For me, ultimately, if "artifacts" are playing into what makes these prints look better, that doesn't automatically invalidate that they are perceived as looking better, especially if the artifacts are in some sense ghosts of actual detail that was captured.

Do you have any idea whether the D800 will have this issue as well? Which DSLRs don't have it?

Best,

Brian
So do you want blurry pixels that are a good approximation to the continous waveform, or sharp (aliased) pixels that have the "right value, but the wrong location"? I think that actually photography lags behind in this discussion, designers of computer fonts seems to be a lot more conscious of the trade-offs, perhaps because theoir "product" consists of 8x8 pixel "images" where one really can afford to obsess over single-pixel details.
http://www.joelonsoftware.com/items/2007/06/12.html

-h
« Last Edit: February 16, 2012, 04:30:57 am by hjulenissen »
Logged

bdosserman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 120
Re: Everything matters, science, etc.
« Reply #19 on: February 16, 2012, 12:23:15 pm »

Hi,

I am certain that there is more to print sharpness/detail than the number of dpi. If you really want to establish good lower bounds based on empiry, you probably dont want to use a general purpose image processing pipeline that includes god knows how many image scalings, an unknown dithering pattern, and unknown ink splatter characteristics. Not that scaling in a non-linear color/intensity representation (such as sRGB) actually gives some (usually small) errors compared to moving the image closer/further away from the viewer.

This is a good point. To me, from a naive point of view this supports the argument that resolution increases beyond what the eye is normally capable of seeing could end up producing perceptible improvements to the final image.

Quote
I think that image perception research have suggested that 1 minute of arc is a good rule of thumb for a normal-good pair of eyes. I presume this is for relatively high-contrast black/white transitions perfectly smoothly/sharp printed on paper. This could be compared to actual prints at a certain size/distance/resolution, and if they are in the same ballpark, I would be satisfied.

Is this for what the eyes can "resolve", or the point at which they can no longer tell the difference between higher resolutions? I could imagine these could be pretty different. [Also, a side comment on "normal eyes", following up on what someone brought up in a different thread: my vision is terrible, but with glasses it corrects to about 20/15. I guess this is a noticeable difference from 20/20]

Quote
Sure. But when do you stop? Is it worthwhile to have an image that prints 1x1 meter, where you can step closer until you reach the close-focus limit of your eyesight and still see no resolution limitations?

No, obviously not. But I'm just pointing out that one should include some leeway for the presumed viewing distance.

Quote
"Megapixels" is a horrible concept. But I do believe that Bayer sensors behave very good for the kind of scenes that photographers tend to shoot, and for viewers that are human. Certainly a lot better than what a simplistic analysis of the red/green/blur sensels would suggest.

This certainly seems to be the case.

Quote
Does it have to be either or? I think that doing both and relating them to each other is a good approach.

Of course not. I find the theoretical discussion interesting, and ideally would like to have it and the empirical discussion developed to the point where both are in agreement. It just bothers me when people present a theoretical analysis as if it should close all further discussion. In general, but especially here where the estimates of what resolutions the eye can readily perceive are varying by factors of four, right in the range that completely changes the answer to the original question.

Quote
Oversampling is a method where you can replace a complex analog filter with a (equally complex but simpler to design) digital filter. Or you might see it (with dithering) as a method to provide many levels at lower sample-rates, when the physical technology is better suited for fewer levels and higher sample-rates.Scaling your typical over-sampled image is not that hard. Scaling an image that has been excessively over-sharpened/aliased through non-Nyquistian sampling is harder.

Can you explain the last sentence? I'm not familiar with the terminology.

Quote
Many image processing algorithms and printer drivers seems to use very naiive image scaling algorithms. Perhaps because most of their customers dont care that much.

It would be interesting to know more about this part of the pipeline, and how much it's affecting the final output quality.

Quote
So do you want blurry pixels that are a good approximation to the continous waveform, or sharp (aliased) pixels that have the "right value, but the wrong location"? I think that actually photography lags behind in this discussion, designers of computer fonts seems to be a lot more conscious of the trade-offs, perhaps because theoir "product" consists of 8x8 pixel "images" where one really can afford to obsess over single-pixel details.
http://www.joelonsoftware.com/items/2007/06/12.html

I think that's exactly why font designers obsess over it (that plus the obviously related point that their individual pixels are a whole lot more visible than those in a photographic print). I agree that these things ought to be considered more seriously (at least, assuming they haven't been; I don't claim to know what people have and haven't thought about), but with the obvious caveat that what makes sense for fonts is probably not what will make sense for photography, given the very different circumstances.

Brian
Logged
Pages: [1] 2   Go Up