How many megapixels do you really have?

[dan_south]

What portion of a Rembrandt contains visible detail? Do art dealers deduct for all of those dark shadows? Imagine

the pathetic histogram on one of those images! :-)

 

Or consider a photo containing bokeh. Those omnipresent bridal portraits taken with an 85 mm f/1.4 lens, for

instance. Should she pay only for the pixels that are actually in focus, i.e. one eye and a few strands of hair? Does it

not matter that the high density sensor lets you blow her up to 24x36 with remarkable detail in that in-focus eye?

[zack_zoll]

 

<blockquote>

<p ><a name="00YTgc"></a><a href="../photodb/user?user_id=2071900">Dan South</a> <a href="../member-status-icons"><img title="Frequent poster" src="../v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 29, 2011; 07:17 p.m.</p>

</blockquote>

 

<blockquote>

<p>Or consider a photo containing bokeh. Those omnipresent bridal portraits taken with an 85 mm f/1.4 lens, for instance. Should she pay only for the pixels that are actually in focus, i.e. one eye and a few strands of hair? Does it not matter that the high density sensor lets you blow her up to 24x36 with remarkable detail in that in-focus eye?</p>

</blockquote>

<p>Dan brings up a good point. While this chart is interesting, and technically correct, it is based on a fundamental misunderstanding of photography. Let me make an analogy:</p>

<p>You are catering a wedding. When you go to set the tables, you put a place setting at every seat - ten per table. Six people sit at one table. This means that four place settings are wasted. In practice though they are not, because four other people might have sat there. Or maybe those six people may have selected different chairs, causing a different four place settings to be wasted.</p>

<p>a 10 MP sensor may indeed only yield 6 MP of detail. I lack the physics knowledge to argue, so I'll just assume it's true for now. Even if you're only getting 6 MP, you don't know where those pixels will be needed until you snap the shot, so having many more allows you to have enough pixel density when and where it is needed. A 10 MP camera at f/5.6 may have the same amount of theoretical pixels as a 20 MP camera at f/1.4. However, the 'relevant' part of the 20 MP image will be much sharper and yield a much crisper poster-sized enlargement than the 10 MP image.</p>

<p>Since photosites do not move, the only way to be sure of having enough detail is to cram as many as you *might* need on there, so you can pick and choose; just like you would set a place setting at every chair, since you don't know who will want to sit where.</p>

<p>Beyond that, it's just a fanciful logistics game; the photographic equivalent of cute wordplay.</p>

 

 

[pete_s.]

<blockquote>

<p>A 10 MP camera at f/5.6 may have the same amount of theoretical pixels as a 20 MP camera at f/1.4. However, the 'relevant' part of the 20 MP image will be much sharper and yield a much crisper poster-sized enlargement than the 10 MP image.</p>

</blockquote>

<p>The resolution figures mentioned above is calculated at the sharpest part of the image (the center). So contrary to what may seem logical at first glance the image from the 20 MP camera will <strong>not</strong> provide a crisper poster sized enlargement when the real resolution is the same. You could take the higher resolution image from the 10MP camera and upsample it to 20 MP and it will have exactly the same amount of detail.</p>

<p>That's why resolution and number of pixels are two different things. You might however have a more beautiful image if it was shoot at f/1.4 though and that is of course something to consider!</p>

<p> </p>

[scott_ferris]

<p>There is are several huge assumptions in all this.</p>

<p>The first is that DxO results/comments are worth the pixels used to represent them, the statment of theirs that confirms they are not is that the best resolving Canon lens they have used is the 85mm f1.8. That is, quite simply, a joke.</p>

<p>The second is for you film fanatics, your presumption is that in the same use you get better focus, more stability etc and hence more megapixels! Piffle.</p>

<p>The third is that you don't actually take pictures and print them. If you did you would know that this is all theoretical nonsense, it is proving a bumble bee can't fly by physics rather than looking in the garden and just watching them fly.</p>

[Brad_]

>>> How many megapixels do you really have?

 

Don't know. Don't care.

 

I make photos. If I had more megapixels, or less, it wouldn't make any difference.

[dan_south]

<p>Actually, Scott, there's another misconception that no one has mentioned yet - the idea that the lens is robbing the sensor of pixels. Put a moderately sharp lens on a 12 MP full frame camera. Then put it on a 24 MP full frame camera. Which one will produce an image with better resolution? Of course, the 24 MP camera. And if they release a 36 MP camera next year and you test it with the same lens, it will produce even more detail.</p>

<p>You could put a sharper lens on each camera and test again, or a weaker lens. Each camera still produces output consistent with the resolution of its sensor and processing algorithms. 24 MP will have more detail than 12 MP. Sharper lenses will provide more detail than other lenses on every camera body, but no pixels are taken away. Otherwise, the only time you would be shooting at the full resolution of your camera - of ANY camera, would be if you were shooting through a theoretically perfect lens, a lens that can never exist in the real world.</p>

<p>Misconception Number 2: We see the pixels that the sensor captures. FALSE! We see the output of the camera's image processing software. Remember, pixels are monochromatic - red, green, or blue. When you magnify your images on the computer screen, do you see red, green, and blue dots? NO! You see dots of different shades and colors. The sensor's pixels have been translated by any number of proprietary blending, smoothing, mapping, and interpolating algorithms to assemble the image that you see. That's why, for instance, a 12 MP D3 has much higher resolution than a 12 MP D2x. The D3's internal processing is far more advanced.</p>

[scott_ferris]

<p><em>"You could put a sharper lens on each camera and test again, or a weaker lens. Each camera still produces output consistent with the resolution of its sensor and processing algorithms."</em></p>

<p>The lens does not rob pixels, but it does rob resolution, and that is what the thread is about. Theoretically a digital images resolution, independent of subject or shooting conditions, can be limited by the number of pixels, the lenses resolving power and the aperture. Put a 10mp or a 21mp sensor behind a beer bottle lens and resolution is zero, the file size might be larger for the larger camera but neither sensor can resolve anything. I have a 35-80 somewhere, it can't resolve a chessboard at three feet. Put a supremely sharp lens in front of a 21mp and a 45mp sensor (not yet available but it will be) at f16 and they will maintain the same resolution, they are both diffraction limited, put them at f1.4 and the 45mp image will have more resolution. Why do you think the new generation fast primes are so expensive? They need to deal with the next generation sensors.</p>

<p>Check out Nyquist limits, they give an understanding of why a lens on a crop camera is tested so much harder than a ff one, not the other way around. The 100mm f2.8 IS macro lens is Nyquist limited on a 1Ds MkIII over much of the sensor for a large part of the aperture range, it is, on that camera with that sensor, effectively, a perfect lens.</p>

<p>I don't understand the point of your second comment though.</p>

[bobatkins]

<p>Why not make to title "How much resolution do you really have", then it would be a lot less confusing, a lot less controversial, a lot more accurate and a lot more useful. As it is I'm wondering if you've been taking lessons from KR. <br />Remember what Disraeli, or possibly Mark Twain (or possibly someone else) said (or would have said if they had read this)....</p>

<p>"<strong><em>There are three kinds of lies: lies, damned lies, and resolution estimates</em></strong>"</p>

<p>BTW do not confuse resolution with image quality or even with sharpness. They are not the same thing and no lens appears to be perfect on any sensor, even if the lens "outresolves" the sensor (which in itself is a meaningless term). MTF at spatial frequencies below the Nyquist limit will determine image quality (and the perceived quality of "sharpness").</p>

[zack_zoll]

 

<blockquote>

<p >Okay ... I just want to make sure I'm understanding you correctly Pete... upsampling is the same as having more megapixels? Does that mean I can get the same enlargement quality with my upsampled D100 than I can get with a D3x? Or are you suggesting that you only use the centre of higher-megapixel cameras and throw out the inconvenient bits around the outside? I'm honestly not sure what you mean, but these are the only things I can get from your post, and they both seem a little silly :( I understand that you add pixels when upsampling, but surely you're not suggesting that pixels you made up in Photoshop are the same as pixels seen by the camera.</p>

</blockquote>

 

 

This whole thread smacks of Schrodinger's Cat. Long story short (google it), Schrodinger came up with a way to mathematically prove that a cat is 50% alive and 50% dead. Einstein said, 'BS. The cat is alive or it isn't.' You might be able to prove that there are all sorts of ways in which a lower-megapixel camera can produce a bigger and better print than a higher-megapixel camera, but doing so requires manipulating variables, stacking the odds, and creating a scenario that frankly does not mirror the way in which one would actually operate the camera; just as Schrodinger's example relies on a complex set of variables that would never happen in order to prove something that only works based on the <em>definition</em> of something, and is totally contradicted by the <em>meaning </em>of that same thing<em>.</em>

<blockquote>

<em> </em>

</blockquote>

To argue that megapixels (or the photosites that produce them) magically dissapear with certain settings, only to pop back up again later, is ridiculous. I have a 1000 x 1000 tif file. It is one megapixel, because it contains exactly 1 million pixels. It is an underexposed image of a black wall, and looks blank. It still contains exactly one million pixels. Thus, it is still one megapixel. It contains zero useful data, but it is a 1 megapixel image with zero data. I'm sure that someone can come up with some math to refute that, or tell me that I have a single 'effective' pixel, but the weird thing is I can still right click on the file, and .... whaddya know? It's still 1000 x 1000 pixels. One megapixel.

<blockquote>

 

 

</blockquote>

And as for the film comments ... film has no megapixels, because film has no pixels. <em>Scanning</em> film produces megapixels, which are produced <em>by the scanner</em>. 50 ISO film perfectly focused and developed has the same number of megapixels as 1600 ISO film, blurry and poorly developed. To say otherwise is just to create a wee contest for film versus digital, and there's already enough of those.

 

[zack_zoll]

<p>Whoops. Massive quote fail. <em>Massive</em>.</p>

[pete_s.]

<blockquote>

<p>Bob Atkins wrote: Why not make to title "How much resolution do you really have", then it would be a lot less confusing, a lot less controversial, a lot more accurate and a lot more useful.</p>

</blockquote>

<p>Yes, Bob that would perhaps been better. It seems the term megapixel is infected.</p>

<p>Also I just turned dxomark's lp/mm @ MTF20 into "real megapixels" which is similar to older imatest versions presenting "ideal megapixels". I thought using megapixels as a unit for resolution would make more sense for non-technical photographers and that having some comparative values would be useful as well.</p>

<p>However it seems that the difference between pixel count and resolution is not apparent to all photographers which makes this thread rather unproductive.</p>

<p>I know the difference between "sharpness" and resolution. And I would like to recommend your <a href="http://www.bobatkins.com/photography/technical/">excellent technical articles on your site</a> for those that haven't seen them!</p>

[User_4136860]

<p>I wish photographers cared more about the content and meaning of their images than the technicality s of their equipment, nobody ever cared what quill pens Shakespeare used, or what sort brushes Van Gough painted with , most of the Worlds greatest photographs were taken with equipment that most photographers these days would consider to be crap.<br>

My answer to the original question "how many mega pixels do you have", is sufficient.</p>

[eric merrill]

<blockquote>

<p>However we could <strong>look at it another way</strong>. 44 lp/mm means that a full frame sensor (36x24mm) needs 44*36mm*2 pixels on the long side and 44*24mm*2 pixels on the short side to show all the detail in the image.</p>

</blockquote>

<p>I would have thought you'd need 4 pixels to show a line pair. Otherwise, you're just showing solid black.</p>

<p>So your example of only needing 6.7 megapixels would really require closer to 28 to capture at the maximum of the lens.</p>

<p>What am I missing?</p>

<p>Eric</p>

[jamie_robertson2]

<p>What a complete and utter waste of time. So now we know that if we:</p>

 

<ol>

<li>don't get perfect focus</li>

<li>introduce camera shake</li>

<li>don't use the optimum lens aperture</li>

</ol>

<p>...we end up with poorer quality photographs. Wow, just the same as when shooting film.</p>

<p>Gee, thanks for the advice *sigh*</p>

[steve_swinehart]

<p>All of this is a moot point. You either take interesting photographs or not. This is truly a waste of time - not megapixels. Here's a real life problem, not a theoretical one. I just did a portrait session for woman who wanted some head shots for her agent. I used my Leica M8 with a 75mm APO Summicron. Unfortunately, neither the lens nor the sensor knew about the theoretical loss of information that I would be sufferering from this alleged "problem."</p>

<p>So, now I'm spending time in an editing program cleaning up the blood vessels in her eyes, individual stray hairs on her forehead, etc. I'm not sure I really need to know how much information I've apparently "lost" at this point, as I don't need any more...really...</p>

[bobatkins]

<p>Not to toss more hand grenades into the fray but....</p>

<p>(1) It's misleading to say the lens A on camera B (16MP) effectively has 8MP "real" resolution because if you actually had an 8MP camera, the resolution would drop.</p>

<p>(2) You have the resolution data in order to calculate the "real MP" so why go through aprocess that just makes things more confusing. It's like having mpg figures for a car, then calculating something like "normalized miles per tankful" and using that as a measure of fuel economy</p>

<p>(3) No matter how good the lens, if you have a higher pixel count sensor, you get higher image quality (issues of noise etc. ignored). I've never yet seen a case where I didn't see a sharper image by going to a higher pixel count sensor. I'll admit the differences are small, often much less than a pixel count might suggest, but they are real (again, assuming everything other the the pixel size and pitch stays the same)</p>

<p>So although you have an interesting idea and a novel way of expressing it, I'm afraid it just doesn't work well and doesn't convey your intended point.</p>

[KenPapai]

<p>Using pro lenses, I take a photo at f/8 with my 10D, 40D, and 7D of the moon, ISO 200.</p>

<p>With each EOS camera I get lots more resolution of the moon directly in line with the MP increases of each: from 6, to 10, and to 18.</p>

<p>Real life - real results - real (advertised) MPs matter when shooting real subjects.</p>

[zack_zoll]

<p ><a name="00YTqO"></a><a href="../photodb/user?user_id=2237443">Pete S.</a> <a href="../member-status-icons"><img title="Frequent poster" src="../v3graphics/member-status-icons/1roll.gif" alt="" /></a>, Mar 30, 2011; 05:26 a.m.</p>

 

 

<blockquote>

<p>However it seems that the difference between pixel count and resolution is not apparent to all photographers which makes this thread rather unproductive.</p>

</blockquote>

<p>The word 'resolution' has two different meanings, which is why the comparison does not work, and why I say this argument is mere wordplay. Resolution refers to the sharpness or 'resolving power' of a camera or lens. Resolution also refers to the number of pixels that make up an image; as in, your screen resolution. This (proper) use of the word makes it the <em>same as pixel count</em>, and the fact that you're trying to call people out for not instinctively understanding which is which makes you part of the problem here. The maddening thing about this 'real megapixels' argument is that it is phrased in such a way that when one meaning of the word 'resolution' is argued with, one can turn around and argue about how the other meaning is correct, ad infinitum.</p>

<p>Because of the confusion in the nomenclature, there is literally no way to prove or disprove this argument. The scientific method tells us that if you can't test a theory to prove its veracity, it is not a legitmate theory; it is a belief. It's no different than saying that Holy Book X is true because it they are the words of God, and we know God exists because Holy Book X says so.</p>

<p>This discussion is a belief. The parameters mean it cannot be disproved, and thus cannot be proved. At this point I've already wasted enough of my life discussing it, so I'm giving up to go take photos. I have a late night photo shoot :)</p>

 

[ilkka]

<p>Following on what Bob Atkins pointed out above. The DxO test data seems to show that Nikon D3, a 12.1Mp camera can resolve 9.7Mp with the ''best lens". D3x is a 24.5Mp body and can resolve 15Mp, presumably with the same lens, coming from the same manufacturer. So more megapixels resolves more. So photographers who want to have more detail, bigger prints, or room to crop, should buy a camera with as many megapixles as possible, while of course also using good lenses. What else is new? </p>

[dan_south]

<p>Scott, my second comment (misconception number 2) is that our digital photos are not the Bayer patterns that our sensors capture, but rather they are the output of a process that uses the Bayer pattern as a raw material. We aren't looking at the sensor's pixels when we look at a digital photo. We look at the result of interpolation and smoothing algorithms that use sensor data as a starting point.</p>

[rashedahmed]

<p>Forget everything. Check what your PS software says in Image Size.</p>

[pete_s.]

<blockquote>

<p>Zack wrote:<br>

The word 'resolution' has two different meanings...</p>

</blockquote>

<p>Hi Zack,</p>

<p>I agree that it can be confusing when the proper term is not used. That is why camera manufacturers do not use "resolution" when they are mentioning how many pixels their sensors are using. The proper term is "number of effective pixels". You can check out the specification of any Canon or Nikon camera if you don't believe me. It's in the <a href="http://www.cipa.jp/english/hyoujunka/kikaku/pdf/DCG-001_E.pdf">CIPA guidelines</a> as well.</p>

<p>And it's not apparent at all that a camera that have a higher megapixel sensor has higher resolution than one with less megapixels. Not when you have full frame, aps and four thirds sensors and you can put different lenses on them as well.</p>

<p>For instance using the best lenses at the optimal aperture does the 18 megapixel Canon 7D have higher resolution than the old 12 megapixel Canon 5D?</p>

<p>The answer is no according to dxomark's measurements.</p>

<p> </p>

[bojepsen1]

<p>I must admit I'm rather with Brad on this one: does it really matter?</p>

[pete_s.]

<p>The original post was posted in the wedding forum and it was from the viewpoint of a wedding shooter I wrote this.</p>

<p>For instance when shooting formal group shots that often are enlarged you usually want a high level of detail on the subject. But many times you work in low light so you have to compromise. If you have enough dof you might shoot a group at f/2.8 or even larger. And I was interested to see how much resolution you would lose depending on the aperture.</p>

<p>Another situation would be when you setup strobes or off camera flash. It's good to know what kind of aperture you have to light towards. It would also be interesting to see how much resolution you lose when at different ISOs. And hand holding with good technique compared to a monopod and to a tripod. And image stabilisation lenses versus non-stabilized.</p>

<p>Anyone who has done resolution testing know it is a lot of work if you want to do it right and not everybody has that amount of time. I've shot a fair number of iso 12233 charts and slanted targets so I know. Even if sample variation will influence the results it's good that dxomark has done a number of tests one can use to evaluate performance without doing all the work.</p>

<p>In hindsight just as Bob Atkins mentioned I should have not have introduced my "real megapixel" unit and just used line width per picture height instead as a measure of resolution. Then this thread would have stayed in the wedding forum instead of ending up here to die :-)</p>

<p> </p>

[scott_ferris]

<p>Dan,</p>

<p> That might be true, the image is a clever reinterpretation of what was exposed. But with regards resolution it is important to realise you do get the full number of pixels, and their luminance, the Bayer patterns are used to determine the pixels colour from the information gleaned from the surrounding pixels. The number of pixels is the number of the pixels, they are not made up or interpolated, only the colour is, and from most peoples experience, it is pretty good at it. A 21mp sensor, has 21 million individual light sensitive receptors.</p>

<p>Brad, Bo,</p>

<p>Generally, no, it doesn't matter. I did once have a terrible problem at a wedding where the guys were wearing kilts, the pattern was the wrong frequency and caused moire with one camera but not with another from the same tripod. How well you understand the technical background of your camera does not make your images better, but understanding limitations can stop them being worse.</p>