What's new

Extra Megapixels Useless?

c.cloudwalker said:
>>>SNIP>>>One cannot create what isn't there in the first place.>>SNIP>>

And the problem is the same, inventing pixels that are just not there. It may become possible some day but I don't see it today.

Well, "inventing pixels that are just not there" is precisely, exactly, and simply, what the demosaicing routine does when it creates the images off of every single image made with every single Bayer-sensor arrayed camera. The images you are looking at today, in 2011, are created with "estimated" color information...and current RAW file converters create the "illusion"of detail and texture above the Nyquist limit...and I say this not to appear dickish, but just to set you on the path of better information...what you think will be possible "someday" is the way digital imaging actually works at the most-basic level.
 
c.cloudwalker said:
>>>SNIP>>>One cannot create what isn't there in the first place.>>SNIP>>

And the problem is the same, inventing pixels that are just not there. It may become possible some day but I don't see it today.

Well, "inventing pixels that are just not there" is precisely, exactly, and simply, what the demosaicing routine does when it creates the images off of every single image made with every single Bayer-sensor arrayed camera. The images you are looking at today, in 2011, are created with "estimated" color information...and current RAW file converters create the "illusion"of detail and texture above the Nyquist limit...and I say this not to appear dickish, but just to set you on the path of better information...what you think will be possible "someday" is the way digital imaging actually works at the most-basic level.

Maybe so but you are talking about within the small size of the sensor.

I was responding to the 1000% comment.
 
According to DxOMark, the best lenses for DSLRs have a resolution of 66-67 lp/mm - which equals to about a maximum circle of confusion of about 15 µm. Using the Diffraction Limit Calculator of http://www.cambridgeincolour.com. The maximum pixels a full frame sensor should have is 24 megapixels and crop sensor should have about 10 megapixels, why do lots of crop sensors have 16-18 megapixels?

Could you give us your calculations for those values. They look wrong, but I'd like to see your calcs.

Common sense will tell you that when the sensor's resolution is higher than the lenses resolution, more MP will look like bicubic upsizing. Just my common sense. I might be wrong.

Remember that the resolution of a system is not the resolution of the lowest-resolving element of the system. It can be reduced below that.

Remember that DxO Mark resolution figures are not for the lens, but for a lens/camera pair. The lens resolution could be higher. Note that the same lens achieves higher resolution on the D3x than on a D3, for example (that alone seems to blow a bit of a hole in your argument).

Best,
Helen
 
According to DxOMark, the best lenses for DSLRs have a resolution of 66-67 lp/mm - which equals to about a maximum circle of confusion of about 15 µm. Using the Diffraction Limit Calculator of http://www.cambridgeincolour.com. The maximum pixels a full frame sensor should have is 24 megapixels and crop sensor should have about 10 megapixels, why do lots of crop sensors have 16-18 megapixels?

Could you give us your calculations for those values. They look wrong, but I'd like to see your calcs.

Yes, it seems wrong. Resolution of 66-67 lp/mm = Circle of confusion of about 15 µm. Agree? Maximum pixel pitch for circle of confusion of about 15 µm = 6 µm. 6 µm pixel pitch for each pixel on 24mmx18mm = about 10MP. Correct? Wrong?

Common sense will tell you that when the sensor's resolution is higher than the lenses resolution, more MP will look like bicubic upsizing. Just my common sense. I might be wrong.

Remember that the resolution of a system is not the resolution of the lowest-resolving element of the system. It can be reduced below that.

Remember that DxO Mark resolution figures are not for the lens, but for a lens/camera pair. The lens resolution could be higher. Note that the same lens achieves higher resolution on the D3x than on a D3, for example (that alone seems to blow a bit of a hole in your argument).

Best,
Helen

Yes. I know that. But when you look closer, why does a lens have a higher resolution per mm in D3x but a lower resolution in D7000. Ain't D7000 have a smaller pixel pitch?
 
The maximum detail a D7000 can "see" is 85 lp/mm. I don't see any lens with that performance.
 
6366804541_7180c3422b_z.jpg

While individual sensor pixels may not resolve a lens, the collection of pixels which make up a single image pixels is twice as large in either dimension. All four pixels are used to describe the average illumination of the CoC at any given 4x4 sensor pixel area, rather that each individual sensor pixel is of higher resolution or not.

http://en.wikipedia.org/wiki/Color_filter_array
Bayer filter - Wikipedia, the free encyclopedia
http://en.wikipedia.org/wiki/Demosaicing
 
Last edited:
The maximum detail a D7000 can "see" is 85 lp/mm. I don't see any lens with that performance.

Not necessarily. You are forgetting the MTF of the AA filter - which is why the D3x can have a higher system resolution than the D7000 with its slightly smaller pixel pitch.

The mentions of a 15 µm c-o-c and the diffraction limit in your calcs was what looked wrong. Going directly from 67 lp/mm to a maximum pixel pitch of about 6 µm is better - though that comes to about 12 MP for a 24 mm x 18 mm sensor, doesn't it? (not that it matters much). You then need to account for the MTF of the AA filter/sensor part of the system, which could be about 0.3 lp/pixel at 20% modulation. As I mentioned before, you need to remember that the DxO figures are not for the lens but for a lens/camera system.
 
In the real world, FOCUSING MISTAKES, and focus shift from wide-open to shooting aperture are simply huge, huge limiting factors on how much detail is resolved. Perhaps the most serious problem with the ultra-high resolution d-slrs and medium format systems is setting a focus point that maximizes resolution. If the focus is off just a little bit, a "24 megapixel" D3x capture can easily,easily have lower resolution than a well-focused D3s photo that is properly focused. Camera motion or vibration are other sharpness killers.

Take a look at this review, about 3/4 of the way down the page, and look at 12MP versus 24 MP captures of the same subject matter:
Nikon D3x Review by Thom Hogan

Actually "achieving" 24 megapixels's worth of image is quite a challenging endeavor...and, a GOOD 12 MP capture can be up-rezzed to look an awful lot like a 24MP capture. If you're shooting for publication and your files are being screened, MP count doesn't much matter as long as the focus is decent and the lens is good...SPorts Illustrated has been running double-trucks since the D1 days (2.7MP) and the Canon 1D days (4.2 MP), and the images as printed in the magazine look quite good no matter what the MP count of the camera was. Same on newsprint...the screen is so coarse and the paper so "meh" that the camera's MP count is almost meaningless.
 
The maximum detail a D7000 can "see" is 85 lp/mm. I don't see any lens with that performance.

Not necessarily. You are forgetting the MTF of the AA filter - which is why the D3x can have a higher system resolution than the D7000 with its slightly smaller pixel pitch.

The mentions of a 15 µm c-o-c and the diffraction limit in your calcs was what looked wrong. Going directly from 67 lp/mm to a maximum pixel pitch of about 6 µm is better - though that comes to about 12 MP for a 24 mm x 18 mm sensor, doesn't it? (not that it matters much). You then need to account for the MTF of the AA filter/sensor part of the system, which could be about 0.3 lp/pixel at 20% modulation. As I mentioned before, you need to remember that the DxO figures are not for the lens but for a lens/camera system.

This gave me another good idea. I think if camera manufacturers created vignettes around each pixel then the AA filter/low pass filter could be eliminated.

EDIT: Lol. Stupid idea that won't work.
 
Last edited:
^^ Lol. I was kind of scratching my head a bit. :greenpbl: << wtf is this guy, anyway?
 
Did I mention I missed the film days when all we had to worry about was pushing a roll of 400asa to 1600asa and hope it held together enough to get a decent print.
 
^^ I dunno. You still had Rodinal v. Microdol and the whole accutance v. resolution debate.
 
I didn't really think much about what we were using, depended on what neg film we were shooting, ilford/kodak, played with different temps, processing colour, how many times to turn the can over while processing. It was all pretty simple stuff. I don't ever remember getting into many discussions with the other photographers about processing. As long as the grain didn't look like golf balls, which it did on occation, and we could squeeze a decent 8x10 out of it the world was right.
 
I didn't really think much about what we were using, depended on what neg film we were shooting, ilford/kodak, played with different temps, processing colour, how many times to turn the can over while processing. It was all pretty simple stuff. I don't ever remember getting into many discussions with the other photographers about processing. As long as the grain didn't look like golf balls, which it did on occation, and we could squeeze a decent 8x10 out of it the world was right.

I recall this era as being filled with HUGE discussions and arguments as to how much water to add to the coffee maker, and how long a pot of coffee could sit before it was deemed "sludge", as well as some pretty bitter arguments about whose job it was to make coffee;perhaps no argument was as bitter,nor as frequent, as the one about HOW MUCH COFFEE was optimal to achieve the desired degree of jangled nerves; darkroom techs/photogs were expected to make coffee, since we knew how to measure water, mix powders, and insert filters, as well as how to handle complicated tasks such as rinsing out the glass carafe without breaking it, as well as handling the nuances of the On/Off switch and the Timer/No Timer settings. Of course, the paste-up crew drank the majority of the coffee as they waited for halftones to come out. Souping pushed rolls of Tri-X to E.I. 3200 meant an ounce of HC-110 syrup per 4-roll tank and about 15 minutes at 76 degrees Farenheit...and all you really had were highlights and mid-tones and zero shadow detail...
 

Most reactions

Back
Top Bottom