PetaPixel

Perceptual Megapixel: Lens Sharpness Boiled Down to a Single Number

Modulation Transfer Function (MTF) charts are a very commonly used tool in the photo industry for measuring and describing how sharp a particular lens is. However, it’s a system that is largely enigmatic to those outside the realm of optic experts and camera gearheads.

For those of you who don’t want to learn how to read MTF charts, camera gear testing service DxOMark has announced a new metric that boils a lens’ sharpness down to a single easy-to-understand-and-compare number: the Perceptual Megapixel.

The system attempts to make sharpness a more intuitive concept by using megapixels to describe it. After all, photographers spend their time working with photographs rather than scientific optical testing equipment.

Can you easily read this MTF chart? If not, then the Perceptual Megapixel is a measurement designed for you

Perceptual Megapixels, P-MPix for short, can be described as the “equivalent” number of megapixels when using a particular lens. Just as lenses offer different equivalent focal lengths when paired with different sensor sizes, sensors can have different equivalent megapixels when paired with lenses of various optical qualities.

If you stick a horrible lens onto a high-megapixel camera, your photographs may pop out with a large megapixel “size”, but the sharpness of the photo may be equivalent to the sharpness of a photo taken with a much smaller sensor and a “perfect” lens.

For example, say you’re shooting with a 24-megapixel camera, but are using a lens rated at 18 P-MPix. This means that the resulting photos are equal in sharpness to an 18-megapixel camera shooting with a optically perfect lens.

What this new system allows for is direct comparisons between camera/lens combos and nice graphics showing how they perform.

Here’s a chart showing the P-MPix values of the original Canon 24-70mm f/2.8 on a 5D Mark II:

Using the system, DxOMark also tested more than 2,500 camera and lens combinations. The conclusion they came to is startling: on average, 45% of a camera’s reported resolution is lost to defects in the camera or lens.

Here’s a chart that shows how to highly regarded lenses stack up P-MPix-wise to the Canon 5D Mark II. Although the DSLR offers 21-megapixels of resolution, the lenses don’t take full advantage of them:

Here’s another chart that shows the drastic difference in perceived sharpness and resolution between a good and bad lens. The affordable Samyang 35mm f/1.4 performs nicely for photographers shooting with the orignal Canon 5D, but on the Canon 7D (which has much smaller pixels) the perceived resolution is less than half of what the camera is capable of:

While it doesn’t seem likely that lens manufacturers will join DxOMark in this new system and include P-MPix numbers in official specs, we’ll likely be seeing the metric referenced around the web in the future as the service uses it to quantify the quality of new glass.

(via DxOMark via CNET)


P.S. Another interesting tidbit: DxOMark writes that though using this new metric, they have confirmed that a 12-megapixel full-frame camera is sharper than an 18-megapixel APS-C camera.


 
 
  • 11

    I just went to the dxo site and yet again realized all those scores for the lenses I mylsef own are mixed up — it’s just against my own tests and also scores elsewhere.. *That score is just garbage* I own some of the best lens from zeiss and canon and canon bodies.

  • beautox

    Also I just looked at the lenses I own with my camera, and the results are simply a crock. It says, for example, that the sharpest lens I have should be the EF 50/1.8 (on a 7D) when in fact it’s simply not. DxO can’t write decent software either…

  • Allan Schroeder

    I don’t like this, it doesn’t specify whether it’s across the whole range of f/stops or how it relates to the corners. MTF data still reigns supreme when you need a specific lens to perform in a specific situation. A number is just that, a number… it has no correlation to real world shooting.

  • wickerprints

    If you have trouble interpreting an MTF chart, you probably have no business using a nonsensical statistic chosen by DxO as a basis of comparison between lenses. Instead, I would advise that you actually, um, take the time to *learn* how to read and interpret the MTF chart itself. After all, there is a reason why the lens manufacturers publish them, rather than some meaningless “perceptual megapixel” number.

    I’ve been saying this for a very long time–DxO’s summary statistics are rubbish. They attempt to distill completely different aspects of imaging performance into a single number and then purport to claim that this number has some kind of relevance to the overall performance of the system. What they publish is no less absurd than a statement like “5.47 + oranges = 2.9 rocket.”

    To critique this specific measure they have created, I need only point out that the most obvious failure is that they have averaged the performance of the lens as a function of image height, f-number, AND spatial frequency. Those are things you simply can’t lump together and summarize with a single number. If I say Lens A has an overall performance of, say, “83%” and Lens B has an overall performance of “78%,” what in the world does that tell ANYONE about how the lenses compare in terms of their center performance, corner performance, wide open, stopped down, or at high and low spatial frequencies? Yes, DxO can provide the measured MTF charts, but then, why bother to devise a bogus summary statistic of that data in the first place? It only serves to mislead the reader.

    For all you know, Lens A could be a 24-70/2.8 shot at 50mm with mediocre resolving power across the entire frame, whereas Lens B could be a 50/1.2 which is understandably soft from f/1.2 – f/1.8 but easily beats the zoom at comparable apertures in the image center. But how would you ever *know* any of this unless you read the MTF? It’s NOT that hard to do.

    DxO basically publishes crap. If they use such nonsensical methods to describe imaging system performance, then it is either incompetence or deception on their part. In either case, that calls into question the validity of the raw data they have measured. So I wouldn’t even trust their raw MTF curves because they’ve shown no understanding of quantitative analysis.

  • http://www.eriklaurikulo.se/ Erik Lauri Kulo

    SCIENCE.

  • http://wemetlastnight.tumblr.com/ Albi Kl

    Could you elaborate on this. I have heard many conflicting reports on the sharpness of the 50/1.8 and honestly don’t know what to believe anymore. What lenses do you have that perform better and does this improvement extend to other bodies apart from the 7D?

  • http://www.shinyphoto.co.uk/ Tim

    It seems to be a function of both the lens *and* the camera, and should be regarded as such at all times.

    Now, back in the days when I scanned film on a V700 flatbed, I always got the impression that there was an unsharpness due to the glass that had some kind of “characteristic radius” to it – whether it was that something that should be a pixel was smeared over 2.5, or what. ISTM DxO’s latest metric might be in the same ballpark – and therefore the most useful thing is whether there’s an equation to convert it into a measure of how much sharpening is required.

  • DafOwen

    I’ve enjoyed DxO results in the past – mostly as it makes Canon fanboys squirm and whinge which is a laugh. But this one does seem too generalised.

    Also – I don’t quite understand how the value would vary across cameras.

    E.g. Nikon 24-70mm 2.8 is:
    11P-Mpix on a D4, but

    14P-Mpix on a D3x

  • Muffist

    well you are an i*i*t…. DXO knows more about lenses or sensors then you internet clown ever will…..

  • Biorman

    well read before you write.. they HAVE tested at different f-stops when you look at the graph.
    but you must be able to undersant the test and get some infos before you write your nonsense.
    so if you want the detaisl you can get the.
    the number is just an overall statement

  • http://www.facebook.com/franklin.weise Franklin Weise

    THAT’s the point. For instance, if one captures landscapes, a lens that performs well at small apertures and that’s sharp in the corners is preferred. But the P-MPix number wouldn’t help much – one still has to check the MFT graph.

  • http://wemetlastnight.tumblr.com/ Albi Kl

    Thanks. Discus/Petapixel seems to act funny with comments that include links but I did manage to read your reply.

    I had considered after market brands to save cost but had been concerned about image quality suffering. It’s’ good to know that even Samyang can compete with some of Canon’s offerings.

  • beautox

    Also some lenses people buy due to the quality of their blur as much as their sharpness. And other things like micro and macro contrast, transition to oof regions, etc

  • F8INTH

    Oh? 9P-Mpix for 70-200 2.8 L IS II? DXO is crap

  • nab111

    Well, according to DxO, my crappy Canon EF-S 18-55mm f/3.5-5.6 IS II is sharper than my 16-35mm f/2.8L or 24-70 2.8…… Really?!?! ….Or am I totally misinterpreting the data?

  • Ivan

    I checked Pentax lenses at DxO and went through most of them, there are a few I actually own knowing exactly what to expect from them… perceptually. I was checking that more complex diagram showing P-MPix as a function of FL and aperture.

    Now: EVERY SINGLE LENS I looked at is sharpest wide open regardless of focal length! Maybe I am reading their charts wrong, or they did something wrong, or Pentax makes lenses that defy the laws of optics, but the green is always at the bottom (wide open), meaning highest P-MPpix. That does NOT match my experience nor Photozone tests. It suggests completely new approach: “for best sharpness, shoot wide open”. Perceptually… I don’t think so.

  • 11

    I wonder how on earth they [dxo] came to become a bench-marking folks?

  • Crusader Ky

    Canon 40mm f/2.8 is THE sharpest budget lens you’ll ever get – dramatically better than the Canon 50mm f/1.8 @ 2.8. See the corner comparison in the review on thedigitalpicture and drop your jaw…

  • JosephRT

    I am assuming that could be because the D3x has larger pixel pitch than the D4

  • http://www.facebook.com/javi.wright Javi Wright

    Completely agree. DxO is just rubbish to be consumed by lazy ignorant people who have to justify their new acquisitions.

  • Love_My_Country

    I feel the camera commandos who are bashing any real sense of realistic real world measurements vs a world black lines and small numbers, need a kindful reminder to chill out. The Nikon 105mm 2.8 is optically superior and MTF superior to that of the sigma 105mm 2.8 hsm. However as I own both as well as a nagging critical wife she sees a difference without even knowing which lense is which. I see a difference. That Mpixel whatever is a real world scenario. On my d7000, the results on dxo prove what I am seeing. Maybe not on other cameras but on a d7000 this is what I see. Secondly it shows that a cannon body and lenses are integrated to work together vs Nikons working against each, something I found owning both. Most importantly I see the trends in my camera body: 85mm is the golden lens for my camera. I don’t care if its all camera bodies, just mine. So keep your pants on and don’t be so pissed you paid a ton of money for a Zeiss and a samayang is better. :)