JackM
Veteran Member
I thought this was the case but then after looking at lens reviews on DxO I'm not so sure. The same lens gets a much worse "sharpness" rating on the 7D2 vs 5DSR.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Can I assume my 7D2 looks like a crop from the 5DS?
- Thread starter JackM
- Start date
Keith Z Leonard
Veteran Member
Don't they test with a "normalized" image size? That would make sense as they would be resampling the 50mp image vs a resampled 20mp 7D2 image.
Slideshow Bob
Senior Member
I guess that depends on whether DXO are measuring sharpness per area, or sharpness per pixel. Also consider that the 5DSr has a cancelled AA filter, so it will have higher acuity than the 7DII. Having looked at 100% crops from the centre of 5DS and 7DII shots, I think the 7DII probably has a slightly stronger AA filter, but it's not much different.
If the pixel pitch is the same, and lens is the same, then the AA filter will be the only differentiator.
Tazz93
Veteran Member
The 5DS/r will have a better per pixel look... however, for most people, the difference is not a factor. The honest truth is, most people make enough mistakes in a capture to mask the difference.
Documentation?
If you are looking at the "Perceptual MP" or P-MP rating on DXOMark, it is a metric of the camera + lens. Since the 5DSR is a 50MP camera, a perfect lens would deliver a P-MP score of 50P-MP. A perfect lens on the 7D2 would deliver a P-MP score of 20MP.
I have both cameras and consider them to be a perfect combination -- 7D2 for action and 5DSR for everything else. To me, they seem very similar on a per pixel basis for any given lens.
Tazz93
Veteran Member
Documentation... I shoot both, but don't take my word for anything... go look for evidence on your own. There are crops all over the "internets", make your own judgement.
Last edited:
J
j900
Guest
Should we deduce that your post above was irrelevant?
J
j900
Guest
They are similar on a per pixel basis but this is not what the DxO metric measures. The lens metric tells yo how much useful megapixels you should get from a body+lens combination. This will favor high mp bodies and large sensor formats for a given pixel pitch, this latter point because the picture is enlarged less to match a given output size.
This metric is sensible to me as is their sensor metric because both are informative on how good your picture is going to be once printed at a normalised output size.
This metric is sensible to me as is their sensor metric because both are informative on how good your picture is going to be once printed at a normalised output size.
Tazz93
Veteran Member
Its as relevant as you want it to be. A basic question was asked, and answer given which can be witnessed for yourself. This site gives you a great tool to compare the differences. Other sites have taken it to another level and equalized pixels on a image crop level, and I have said as an owner of both that I agree with all the resources. If you need more, that's fine... but why are you asking me to convince you? Believe what you want, it will make me no difference. Remember, I'm not asking for your approval of my opinion, I'm just sharing my experiences.
Stollen1234
Leading Member
you dont have to convince people..thats very true..but when you say:
The 5DS/r will have a better per pixel look.
it sounds like law or a fact..here as in science you need to provide evidence or proofs...
The 5DS/r will have a better per pixel look.
it sounds like law or a fact..here as in science you need to provide evidence or proofs...
John Sheehy
Forum Pro
I doubt that any lens could score 20 P-MP on a 7D2; the AA filter would prevent that score. I suspect that a 20MP APS-C could easily do 20 P-MP with a halfway-decent lens if the photosites were tiny (very low fill factor) and had no microlenses or AA filter.
However, scoring as many of DxO's "perceptual MPs" as the camera has MPs is not a good thing; it means tremendous aliasing, and lots of false, Nyquist mirror image frequencies. The best matches are probably where P-MPs are about 40% of MPs.
We like shortcuts to acuity, but to get "20 P-MP" without artifacts probably requires at least 50MP, and 50 "P-MP", at least 125 MP.
The idea of perceptual megapixels is actually nonsense, IMO. It almost implies that photographic captures should be as sharp as computer-generated text and lines and dots, which have aliasing-level acuity, but don't necessarily look aliased because the letters are formed from pixels. Photographed letters at the same size in pixels with the same acuity are extremely aliased and garbled, due to random alignment of letters and pixels.
John Sheehy
Forum Pro
There's no such thing as "useful pixels" vs "useless pixels". If a resolution test gives 12 "perceptual megapixels" for a 20MP camera and a lens, there are no 8 million pixels that you can remove from the image because they are redundant. There are no redundant pixels. In fact, if a 20MP camera scores "12 P-MP" with a certain lens, the camera does not have enough pixels for the lens, I guarantee.
mburns
Forum Enthusiast
I think that you have your argument on aliasing reversed. Digitally generated text and images can be out of alignment with the display, and, regardless of proper alignment, require anti-aliasing before display.
But an ideal digital sensor, with no Bayer color filter array, an ideal micro-lens array, and no spurious sharpening and clarity added by post processing, produces images that are naturally anti-aliased.
But an ideal digital sensor, with no Bayer color filter array, an ideal micro-lens array, and no spurious sharpening and clarity added by post processing, produces images that are naturally anti-aliased.
You should probably read this
Anti-aliasing filter - Wikipedia
He was talking about the optical AA filter built into the camera above the sensor. That AA filter simply makes an effort to smooth edges by blurring the image (it is a low pass filter). It isn't very sophisticated.
Whenever you sample on quanta like pixel sites you are most certainly creating aliasing. This is totally independent of the display medium. The quanitization is part of the captured image.
The 5DS has the usual AA optical filter built in, the 5DSR does not.
Tazz93
Veteran Member
That's because it is "fact" to me... Judge for yourself.Stollen1234 said:
Please forgive the aperture difference, I allowed the metering systems to determine the exposure for themselves. For this purpose I believe the difference is a non factor.
As stated previously, the difference is there, but just not big enough for most to care about. However, I suspect the 5DSr will be even more noticeable.
5DS was shot with a 2 second Mirror Lockup
7D2 was shot with a 1 second Mirror Lockup
Last edited:
mburns
Forum Enthusiast
The post that I replied to does not actually reference an AA filter implicitly or explicitly. And an AA filter for Bayer array sensors is not a primitive device at all.
The article that you linked to is far too short for a proper discussion of natural anti-aliasing by an ideal digital sensor with no Bayer color filter array.
The article that you linked to is far too short for a proper discussion of natural anti-aliasing by an ideal digital sensor with no Bayer color filter array.
John Sheehy
Forum Pro
No; I was referring to when that isn't done. We've all seen non-anti-aliased text and graphics on our screens, every day, and that sets the bar for our maximum expected sharpness.
Naturally non-aliased imaging without AA filters is possible with a Bayer CFA, too; it just takes a lot of pixel density and/or very soft optics at 100% pixel view. There is no need at all for the same color capture characteristics at every spatial sampling location, like a monochrome or Foveon-like sensor. Sufficiently magnified, every apparently abrupt black/white edge transient projected by a lens is actually a slope, not a true cliff, and if the color sampling points are frequent enough to fully define that curvy slope, it doesn't matter one bit where they are relative to the other colors' samples. Sampling three colors or grayscale at every pixel location requires less pixel density for proper, non-aliased sampling, but is not necessary for it.
JackM
Veteran Member
I'm having trouble finding these crops. I'm open to your suggestion, but I'd really be interested to see what you're seeing. Do you have some links handy?
Keyboard shortcuts
- f
- Forum