DxOMark "Perceptual Megapixel"

[BruinBear]

Anyone else see this article?
It makes it seem that current sensors are out-resolving lenses which I dont believe is the case.

http://www.dxomark.com/index.php/Publications/DxOMark-Reviews/Looking-for-new-photo-gear-DxOMark-s-Perceptual-Megapixel-can-help-you

 

[PackLight]

More makebelieve numbers from DxO?

 

[Dasengle]

Anyone else see this article?
It makes it seem that current sensors are out-resolving lenses which I dont believe is the case.

I didn't read the link. But I can believe it. It depends on the lens of course. A 'sharp' lens at lensrentals tests at around 25 line pairs per mm (note that this assumes that all imperfection lies with the lens). That's a roughly 20 micron sized line. So the lens can resolve 20 micron features. Pixel pitch on a d800 is something like 5 micron. The bayer filter musses things up a bit. But its close, a twenty micron feature covers more than two pixels. The sensor doesn't out resolve this lens by much, but going forward new sensors will.

I vote "plausible."

 

[Drizzt321]

Paging Neuro. Doctor Neuroanatomist, please post to this thread.

Overall, it looks like there is something to the 'perceived' resolution, and that the problem they are stating, that MTF charts are not what the vast majority of photographers read, seems fine. But I think they need to publish (if it wasn't in the article, I may have missed something) how they came up with the numbers. Give us the methodology, and then we (meaning not me, someone like neuro probably, or Roger at LensRentals.com) can run their own numbers and thoughts and see if DxOMark's number makes sense.

 

[RLPhoto]

If you have a duff lens, well its not going to get your sensors worth but less. If you have a fantastic lens, Its going to resolve your full sensors megapickel value.

 

[Botts]

Paging Neuro. Doctor Neuroanatomist, please post to this thread.

Overall, it looks like there is something to the 'perceived' resolution, and that the problem they are stating, that MTF charts are not what the vast majority of photographers read, seems fine. But I think they need to publish (if it wasn't in the article, I may have missed something) how they came up with the numbers. Give us the methodology, and then we (meaning not me, someone like neuro probably, or Roger at LensRentals.com) can run their own numbers and thoughts and see if DxOMark's number makes sense.

I was fairly certain that Roger at LR had concluded the D800 out resolved quite a good deal of Nikon lenses.
Roger's Article on the D800 and Lens Choice

The D800′s ultra-high resolution sensor can provide amazing detail. Not every lens (in fact, not most lenses) are going to be able to give you the maximum resolution the camera is capable of. This list isn’t about great lenses, it’s about what lenses can wring the most resolution out of a D800 when you need every ounce of resolution. Maybe you have a two page magazine spread to shoot, or more likely you just want to post your pics on a Canon forum to rile everyone up.

 

[Drizzt321]

Paging Neuro. Doctor Neuroanatomist, please post to this thread.

Overall, it looks like there is something to the 'perceived' resolution, and that the problem they are stating, that MTF charts are not what the vast majority of photographers read, seems fine. But I think they need to publish (if it wasn't in the article, I may have missed something) how they came up with the numbers. Give us the methodology, and then we (meaning not me, someone like neuro probably, or Roger at LensRentals.com) can run their own numbers and thoughts and see if DxOMark's number makes sense.

I was fairly certain that Roger at LR had concluded the D800 out resolved quite a good deal of Nikon lenses.
Roger's Article on the D800 and Lens Choice

The D800′s ultra-high resolution sensor can provide amazing detail. Not every lens (in fact, not most lenses) are going to be able to give you the maximum resolution the camera is capable of. This list isn’t about great lenses, it’s about what lenses can wring the most resolution out of a D800 when you need every ounce of resolution. Maybe you have a two page magazine spread to shoot, or more likely you just want to post your pics on a Canon forum to rile everyone up.

Ah, I think that was before I started reading his blog regularly. Thanks!

 

[Loswr]

[quote author=DxOMark]
Can you define MTF? Do you know if an MTF20% of 50lp/mm is better or worse than an MTF 50% of 30 lp/mm? And when reviewing an MTF chart, can you distinguish which curve is best? The short answer is probably no.
[/quote]

Yes, yes, and yes.

[quote author=DxOMark]
Perceptual MPix:
* Describes resolution with a single number,
* Correlates with the way the human vision perceives resolution
* Uses a unit that is well-known to photographers — the megapixel.
[/quote]

Agree with the above that we need more details about the methodology behind this 'new' score. Because the first two bullet points sound like the Subjective Quality Factor developed by E. Granger at Kodak in the early 1970s. PopPhoto uses SQF for their lens tests, and Imatest reports it for camera+lens, so it could be normalized to the MP count of the sensor (the folks at DxOMark know a little about normalizing ).

 

[Drizzt321]

[quote author=DxOMark]
Can you define MTF? Do you know if an MTF20% of 50lp/mm is better or worse than an MTF 50% of 30 lp/mm? And when reviewing an MTF chart, can you distinguish which curve is best? The short answer is probably no.

Yes, yes, and yes.

[/quote]

Yea, but can the rest of us? I keep re-reading the same couple of articles now and then on MTF, and I sorta, kinda can describe the general purpose, but have a hard time actually getting meaningful information out of MTF charts still

 

[2n10]

So does this mean that all Nikon lenses will be rated better than Canon lenses? :

 

[Area256]

Seems reasonable. I know most of my lenses are being out resolved by the 20MP sensor in the 6D - and that's only in relation to the 100mm f/2.8L which seems to have the best resolution of the lot. However even it could be under resolving compared to what the sensor can do.

Another bit of evidence, if you look at most review sites, you'll find one or two lenses that reach very high numbers, and the rest won't come close to that - so I'd assume based on that, that yes most lenses are being out resolved by the sensor.

Hence I never really did get the D800, unless you are also going shoot everything on a tripod with the best possible glass, it's overkill resulting in wasted space for most other applications. Although great for landscape and product people who own awesome glass and shoot on tripods already.

 

[FatDaddyJones]

DxO said that as a rule of thumb, that "A 12 MP full-format camera is sharper than a 18 MP APS." I not denying the fact, but can someone explain to me why that is?

 

[verysimplejason]

DxO said that as a rule of thumb, that "A 12 MP full-format camera is sharper than a 18 MP APS." I not denying the fact, but can someone explain to me why that is?

It maybe due to the fact that a IQ is very much a factor of pixel quality and not just the pixel quantity alone.

 

[FatDaddyJones]

I've read on other photo sites that the original 5D could resolve sharper images at 12mp than the 7D can at 18mp. The 7D is a newer generation camera, and certainly higher in its MP count. Would it have to do with pixel size perhaps, since the size would have to be much smaller to fit even the same amount of pixels on a crop sensor than on a full frame sensor? If the actual sharpness is greater on the lower MP full frame sensor, then "perceptual megapixel" should be a valid empirical rating scheme and not just a subjective rating.

Thanks Jason for pointing that out. I'd still like a more technical explanation of why this is true.

 

[sandymandy]

The closer u fit the pixels on the sensor near each other the easier noise will start to appear. And any kind of noise even if so subtle means image loss.
In worst case u have a P&S camera with a really tiny sensor with many pixels so close to each other that it will be outresolved by the same sensor with less megapixels.
Some people even say the 1100D has the best IQ in the APS-C line due to this. It just got 12MP but "perceptual MP" is more like 16MP.
But i think this is really not an issue on APS-C or FF sensors. APS-C would be like 60megapixels or such before the pixels come too close to each other.