Fuji GW690III = How many megapixels ?

<p>crop from 5000 pixel high file.</p><div></div>

<p>I'm seeing a lot of content drift away from the original question of the resolution a modern DSLR would need to compete with a 6 x 9 film camera for image quality. Many posts are, instead, about file sizes generated by <em>scanning </em>a 6 X 9 film.</p>

<p>That's fine.</p>

<p>But I think we should be aware that there are really two discussions going on. And that the MP figures reported for scanning film are a lot higher than reasonable estimates for how much a DSLR really needs. That's because it's easy to get such high numbers when scanning. As this thread moves along, we ought to be aware of the direction in which the thread is going, and how much our answers have to do with the original question.</p>

<p> </p>

And also be aware that <i>"the MP figures reported for scanning film are a lot higher than reasonable estimates for how much a DSLR really needs"</i> is not a neutral observation, but one of usual the positions in the DSLRs-are-all-you-need 'debate' that has been raging ever since the first 2 MP DSLRs appeared.<br>(And with that, we should also be aware that, since that claim has indeed been made since those first 2 MP cameras appeared, it's awfully hard to know when such a debate will change from the usual reciting of the DigiDo Mantra into something that has to be given serious attention.)<br><br>The thing is, Rob, that while scanning can indeed produce high pixelcounts, what matters is whether those pixelcounts represent something more than just the number of sensor elements on the scan-bar. Showing and comparing results trying to answer that, and is not topic drift. If we would find the number of scanned pixels per area unit where the transition from true info to empty pixels would occur, we will also get close to knowing (there are other issues to deal with too) how many MP a DSLR would need to equal the results produced by 6x9 film cameras.

<p>Then in order to learn something from scanning 6 x 9 films that can help answer the question of how large a pixel count a DSLR needs to equal a 6 x 9, we might say that we need the largest pixel count that still reveals additional detail in a very fine-grain film, such as Velvia. Ray Butler, above, suggests that 35 to 40MP ought to do it. And he bases this on an experiment, not just on armchair analysis.</p>

<p>I can only think of one thing wrong with that approach. Scanning a film combines two media that are both resolution-limited. I believe that the combined resolution of two media is given by:</p>

<p>(R1 x R2)/(R1 + R2) <br /> where R1 is the resolution of medium 1 (in this case, the digital process);<br /> and R2 is the resolution of medium 2 (the film, in this case)</p>

<p>So if for example, the film can resolve (say) 100 lines per millimeter, and the digital process can resolve 100 lines per millimeter, then:</p>

<p>(100 x 100) / (100 + 100) = 10,000/200 = 100/2 = 50 lines per millimeter.</p>

<p>This would suggest that no matter what resolution you need to get the most out of a 6 x 9 film (35 to 40 MP in Ray's experiment), you have not extracted the full detail in that film. This is because the combined resolution of film and the digital copy of that film must always be lower than that of the film alone. You can come as close as possible to extracting the maximum detail if the digital process has much higher resolution than the film. For example, if the film resolves 100 lines per millimeter, and the digital process resolves 1000 lines per millimeter, then we have:</p>

<p>(100 x 1000)/(100 + 1000) = 100,000/1100 = 90.9 lines per millimeter.</p>

<p>By selecting a digital resolution ten times that of the film's we get a result that is reasonably close to the film's 100 l/mm. But that is misleading, because you don't need 1000 lines per millimeter to <em>equal </em>the film; only to copy the detail in a picture already taken <em>with</em> that given film. So does this method (digitizing a piece of film) really address the original question of the resolution of the DSLR needed to equal the film? I think not. I don't think we can answer the question in this way, owing to what I presented above. The formula I offered for combined resolution needs to be reviewed for relevance by others who are more knowledgeable than I in this area. I think it's right, but I will be interested to see what others say.</p>

And someone else gave different, non-armchair figures.<br><br>The calculations for system resolution do not apply here at all. You are using a device with a (limited) ability to discern thing's sizes to look at the size of things. The result is not a "system" result.<br>Or can you explain why these calculations would apply?<br>Simplified (not taking things like shape, adhesion and what not into account): if you use a sieve with a 1 mm mesh, you can be sure that things smaller than that will fall through, things bigger will not. Reduce the mesh size until nothing comes through, and you will know the size of the smallest thingies you are sifting.<br>If you want to see at what level the resolved detail in an image peters out, all you have to do is look at it with something able to see detail smaller than the smallest detail present in the thing you are looking at. If nothing more reveals itself when looking at even smaller detail you will know what you want to know.<br>Scanning with a 1000 lp/mm resolution will be fine enough, of course, to see that detail in a given film image is not finer than 100 lp/mm. It does not reduce that 100 lp/mm to 90 lp/mm. It does not reduce your inspection instrument's ability to see detail at the 100 lp/mm level.

<blockquote>

<p>The calculations for system resolution do not apply here at all. You are using a device with a (limited) ability to discern thing's sizes to look at the size of things. The result is not a "system" result.<br />Or can you explain why these calculations would apply?</p>

 

</blockquote>

<p>I'd like to see a rigorous explanation to support the position that the combination of film and digital is not a system. Remember, the object is to determine the digital resolution needed to capture the same scene at the same level of detail as a 6 x 9 film can do. If we take the picture with a digital camera, that's not a system (for our present purposes). If we take the picture with a film camera, that's not a system. (Or, if you prefer, it is not a cascading of two systems.) But when we take a digital "picture" (scan) of an analog picture, now we have a system (or system-on-system) whose performance is the composite of both elements of the system; and will be inferior to either one alone.</p>

<p>If we want to know how many megapixels are beneficial to get the most out of a piece of film, or to measure the size of the smallest elements on that film, then certainly the method of successive sieves you suggested above is not merely appropriate; it is an elegant solution! But if we want to know what it takes to capture the <em>original scene</em> with a digital camera, <em>without </em>an analog intermediary (because this was not the original question), we'll have to take the picture twice: once with a digital camera, and once with that (now familiar) 6 x 9. And then take a look. We are certainly on good ground to attempt a theoretical estimate by calculation; but no theory is in good standing until there has been some experimental verification.</p>

<p>If I had a 6 x 9 camera, and a digital camera that had a whole bunch of pixels,I would do it myself. But my D700 has only 12MP, and my only MF is a Hasselblad, which is 6 x 6. I'd welcome some more examples from anyone who may have such examples on hand, or can do the experiment.</p>

<p>A final thought: it is possible that we are asking a bad question that has no correct answer. One might ask how high the sampling rate has to be before a CD can sound like a vinyl record. But it just might be that a CD is not going to sound like a vinyl record, no matter what we do to it; and a digital picture might never look just like an analog one, no matter what. (Just a moment of doubt.)</p>

The explanation, Rob, is that you are not using the scanner to image the detail as present in an original scene through the medium of a lens and film, or anything like that. We are not trying to figure out the ratio between that input's (the scene's) detail and how much there is left on it on film, needing to sum the limiting effects of lens and film through such calculations.<br>Though <i>"the object is to determine the digital resolution needed to capture the same scene at the same level of detail as a 6 x 9 film can do"</i> all we are doing is to zoom in on a bit of film to see what scan resolution would still make sense, and what (higher) scan resolution would only produce more pixels but not reveal more detail/differences present in the subject (the bit of film). The only mediator, the only limiting factor is the scanner. Not the bit to be examined, i.e. the film. No system.<br><br>Should we establish that at a scan resolution of, say, 6000 ppi the extra pixels are (apart from differences caused by noise, etc.) the pixes are mere duplicates of what we alraedy had at lower scan resolutions, i.e. not reveal more detail, we could say that that would be already beyond the MP limit of what a bit of film can deliver. Going backwards we could establish the exact MP number that corresponds with the 'true' image content of the film image.<br>And that would be part of the object as stated. (All else involved, such as lenses, taking technique, etc. assumed to be equal, of course.)<br>The other part is to make people understand that - thanks to Bayer patterns, anti-alias filters (which in itself is an alias: they used to be called soft focus filters) and the like - a, say, 22 MP camera does not capture 22 MP worth of true image content. So if a bit of film would show to hold only 22 MP, a 22 MP camera would not be enough to equal it.<br><br>I don't think it is a bad question, without an answer. Just that the history of it, which it drags along with it, is not a good one.<br>Like i said before (and it's not a joke, not an exageration, or anything like that), ever since the first 2 MP cameras were thrown at us, people have been claiming they were as good as film.<br>And ever since then, people have vehemently defended the 'truth' of that claim, using all sorts of calculations and concepts borrowed from elsewhere, whether they apply or not, and redefined or invented terms (like that anti-alias filter thing, or telecentric, or - a personal favourite - optimized for digital) to shroud inconvenient truths or make them sound less bad.<br>That has all become a bit of a tradition, with people repeating things religiously instead of thinking about what it's all about.<br>And that makes it a difficult question. But not one that cannot be answered.<br><br>But of course there's so much more than pixel count. Whether we like film or digital better is something that really can't be determined by measurements and/or calculations. There are still answers though. Correct ones too. But what is correct, and why to each his own

<p>Well, I think we have both said what there is to say about our respective approaches to this. I'd like to get both our approaches "peer reviewed," so to speak, by sitting back now and waiting to see what others who have given thought to this matter have to say. It seems to me that I want to answer the original question by comparing pictures taken independently with film, and then again with digital; while Q.G. wants to use the digital process as a an instrument to measure what film can do, when expressed in digital terms.</p>

<p>Maybe both ideas have merit. Let's see what others say.</p>

Yes, let's.<br><br>But not before we put one thing straight: <i>"comparing pictures taken independently with film, and then again with digital"</i> as happens here, and other places on the internet, <i><b>is</b></i> using <i>"the digital process as a an instrument to measure what film can do, when expressed in digital terms"</i>.<br><br>The tricky bit in the original question (<i>"how many mega-pixels in a modern DSLR do you think would give the equivalent quality that you get from [film]"</i>) is in the "equivalent quality" part of it. What would that mean exactly?<br>The original question suggests that matching the quality is a matter of the right number of megapixel. And takens as such, using <i>"the digital process as a an instrument to measure what film can do, when expressed in digital terms"</i> is the <i><b>only</b></i> way to settle the question.

<blockquote>

<p>And takens as such, using <em>"the digital process as a an instrument to measure what film can do, when expressed in digital terms"</em> is the <em><strong>only</strong></em> way to settle the question.</p>

</blockquote>

<p>I don't agree with this (and you are welcome to disagree with my disagreement!). For me, photography is about prints. The only valid way to make a comparison is to compare a digitally capture image printed digitally and a film derived image printed optically. Otherwise it's just a digital vs. scanner comparison with the scanner being the weakest link.</p>

You know, i don't disagree, Steve.<br>It's a perfectly valid way to fill in what "equivalent quality" could mean: looks as good as in a print.<br>And for that we need indeed not worry about the number of MP you could eke out of film (which - i believe - is a rather difficult question anyway, if you do not just look at the "resolved and recorded information" content of an image).<br>So i was <i><b>wrong</b></i>.

<p>A couple of things possibly worth reading:<br>

http://toyotadesigner.wordpress.com/tag/resolution-mf-slides/<br>

2007-04-vitale-filmgrain_resolution.pdf</p>

<p>I've done an extensive test of the Mamiya 7 vs various digital - Canon 5Dmk2, Nikon D3X, Phase IQ180 - and the results showed that in terms of clear resolution, the Mamiya 7 resolved more lines than the IQ180 but when you looked at overall rendering it was somewhere between the DSLR's and the IQ180. From this I would estimate that you would need about 40-50 megapixels to emulate T-Max or Velvia 100F and about 30-40 to emulate Portra 160/400.<br>

http://static.timparkin.co.uk/static/tmp/cameratest-2/800px.html<br>

The one thing that came out of this is that digital files can take more sharpening and can look sharper than film but at a certain level of enlargement they start to look wrong (about 240 dpi on close inspection) whereas film has fine detail beyond this. So for large enlargement factors film can look better as long as you don't mind grain. Adding grain to digital helps to a point but it still breaks down at a certain level. We saw this with the IQ180 which looked sharper than 5x4 until you got to 40x50 inch prints at which point 5x4 looked better. <br>

Here's a direct comparison of the IQ180, Portra 160, T-Max and a Nikon D3X<br>

http://bit.ly/Nwzvzp</p>