Megapixels, megapixels and again megapixels

[JimAscher]

Okay, I'll take the bait... Of course I don't still have my 5mp camera - that's not the point. My enjoyment of a photograph is indepependent of the sensor size. While I have a penchant for detail and gain personal enjoyment out of the level I detail I can now achieve, my photographs are not "better" because of the number of pixels present. They are just as "fuzzy" now as before, but with more detail. 

I'll take the liberty of jumping in here to add the following (the relevancy being arguable):  I am leaving for my annual trip to England and France next week (to visit family and/or friends), and this time I've elected to take with me only one of my (too?) many cameras, of multi MP capacities.  That camera is my recently acquired six MP Epson RD1.  After a few months of taking pictures with it I am much taken with the intrinsic(?) emotional quality of its renditions. Sean Reid years ago in one his many essays on the RD1 and the Leica M8 I recall ventured his opinion that the images he'd found from his RD1 were comparable to those he'd experienced with Kodak Tri-X in his film days.  (A film that I'd also used extensively in my own film days.) Thus, the pics I've been taking with my RD1 definitely resonate with me.  They have the character I've been (perhaps subconsciously) striving for digitally ever since.  Can't wait to get to Europe again.  (Plus seeing our rapidly growing teen-aged granddaughter.)

[ErikKaffehr]

Hi,

Some reflections

• On my wall:
  
  I have something like a dozen prints on my wall. All images are printed at A2 size or larger.
  
  - Two images are scanned from 67 slides
  - One is from a 6 MP camera back in 2006. Doesn't hold up to close scrutiny but very nice.
  - One is from a 10 MP camera printed at 70x100 cm (27" x 39"). It is quite OK if viewed at 80 cm or longer.
  - One is from a 16 MP APS-C camera. That one was also shot on 24 MP full frame but the APS-C won out this time. *)
  - Three are stitched on 10, 12, and 24 MP cameras
  - The rest is 24 MP
  
  
• Visual difference
  - Essentially, I see diminishing returns in printing A2 size (16" x 23") past 12 MP
  - Perfectly good A2-prints can be made from 12 MP
  - But, fine details may be lacking from 12 MP. I have an image from Oxbow Bend in Grand Teton NP with some birds swimming around. On the print they look as dust not as birds. 24 MP would have been helpful.
  - I cannot see differences between my 39 MP Phase One back and 24 MP Sony at A2-size. The differences are there and can be seen on the print using a loupe. At A1 and up the difference is visible with the naked eyed. Perceived differences may vary dependent on vision.
  - Printing large (say 27" x 39"), I would say is beyond the limit of 24 MP. Good post processing may still yield very good images.
  
  
• Detail rendition
  The main reason I see for increasing megapixels is correct rendition of detail. If the lens "outresolves" the sensor, fake detail will be generated. So lens resolution is wasted and turned into fake detail.
  
  Check this image (6.8 microns, 18MP on full frame, no OLP filter):
  
  Note how pixelated the text on the sign is. Also we have colour aliasing artefacts. The colour artefacts can be eliminated by and large using an OLP-filter, but that reduces fine detail contrast.
  
  The image below was shot under similar conditions with an APS-C sensor using 3.8 micron sensor, corresponding to 54 MP on full frame. Same focal length, different lens.
  
  
  Here the road sign is highly readable, and no artefacts are visible. The image is a bit softer and more noisy.
  
  Now check the above image downscaled to the same size as the 6.8 micron picture:
  
  
  The resulting image is much cleaner than the 6.8 micron image we started with.
  
• A test chart sample illustrates this pretty well
  
  
  The small pixel camera at top reproduces the "Norman Koren" test target pretty well. It has some artefacts but those are well suppressed.
  The large pixel camera at bottom has a lot of artefacts. High frequency details are "reflected" into fake low frequency details. Intensive colour moiré is seen.
  
  A couple of observations:
  
  - Top row is sinusoidal while second one is "square wave", the square wave has a lot of high frequency components so it shows more aliasing and at lower frequencies than the sinusoidal.
  - The P45+ shows aliasing that is not "symmetrical" about Nyquist. That is almost certainly due to slightly wrong distance chosen in the experimental setup.
  
  
• Good lenses deserve good sensors
  
  The images above indicate that large pixels don't make good lenses justice. They don't resolve true detail and generate fake detail. The fake detail can good look good, unless it is an obvious artefact.
  
  Any decent lens will resolve detail to say 200 lp/mm with low contrast if used at optimal aperture. So buying a camera with a good sensor can extract good detail from good quality lenses, and upgrading the sensor will improve all lenses.
  
  Buying a high quality lens and put it on a low resolution sensor is wasting it's capabilities or even turning it into an aliasing machine.
  
  So, I would say, that if high resolution is not needed, which is probably the case with if printing up to A2-size, the best route may be to buy an APS-C or 4/3-system with high quality lenses. That is also what Ctein indirectly said in his latest interview with Michael Reichmann. He said that 4/3 is good enough for A2 and can do anything he could do with his Pentax 67 on film. Another frequent poster on these threads James Russel (BC) said many times that the Panasonic GH3 is one of his main cameras and good enough for all work.
  
  
• Low light work
  
  Larger sensors make sense for low light work, and larger pixels make some sense in this area. But, large pixels need strong AA-filtering to avoid aliasing. On the other hand, low light work without tripod often fails to achieve the levels of sharpness where pixel size really matters. Lenses that perform well at large apertures are expensive and may be worth the price for available light.
  

Best regards
Erik

[luxborealis]

When you say - "I have a penchant for detail and gain personal enjoyment out of the level I detail I can now achieve" - that suggests your enjoyment of your own photographs is not independent of the sensor size.

Yes, I thought you would pick up on the contradiction.

Could the 5mp photographs be improved if shot at 36mp? They would have more detail, and because of my personal attraction to detail I would probably like them more. Would I have a deeper emotional attachment to the photographs? I'm not sure I would.

Furthermore, detail is not the quality that purchasers of these photographs have been looking at. They buy the photographs because of their overall emotional impact. Is lack of detail sending some potential buyers away, that I couldn't say. As I alluded to in my first post, the pixel peepers are typically other photographers. A couple of them have even remarked on the lack of details - but they weren't looking at the photograph, only the details; IOW they, too, missed the point. But that's okay. We are all blinded by our assumptions.

I've attached digital versions of two of the 5mp photographs. The sheep in Yorkshire are only impressions of sheep due to the lack of detail. The photographers who look at it are upset that they can't count the legs or tell the quality of the wool. Most others become emotionally attached to the overall scene - which is the point of the photograph.

The 45" canvas of Kilimanjaro began life as a 2560x1929 file cropped to 2560x1008. It was up-sampled to 8225x3240, carefully sharpened, then printed to canvas. The uprezzing did not add detail, but given where the canvas is placed, on a wall where people can't get close, it still never fails to draw attention (and not for its lack of detail!).

What I find interesting is how often we confuse detail with quality. It's not unlike a scientist who loses sight of the big picture because their head is in the details. My experience when I took my BSc in Zoology was just that: there were those who got all wrapped up in the minute details of the bacteria living in the guts of various creatures - important research, yes. But I gravitated to the ecologists who were painting the bigger picture of interactions. After my BSc, I took a degree in Geography with (not surprisingly) a particular interest in biogeography and interactions with the human world. Now I was in my element - talk about big picture stuff. Yes, the details are still important, but in a relative sense, those details become relatively less important within the larger perspective.

[Isaac]

Would I have a deeper emotional attachment to the photographs? I'm not sure I would.

Perhaps the emotional attachment is to particular times and places in your own life. Perhaps without that personal attachment more is required of the photographs.

(I took some photographs at a wedding last weekend and I expect a couple of those photographs will be treasured: treasured in-spite of the all too evident faults; treasured because of the relationships they depict.)

The 45" canvas of Kilimanjaro… it still never fails to draw attention (and not for its lack of detail!).

It's nearly 4 foot across -- a 4 foot blank canvas would draw attention.

[Isaac]

In a ideal wold, sensor resolution would match lens resolution. … A good lens stopped down to f/8 will resolve something like 200 lp/mm, albeit at very low contrast. An excellent lens at f/5.6 will perhaps reach 400 lp/mm.

If I did the arithmetic correctly, the Sony APS-C sensor in my camera are something like 105 lp/mm and the bottom of the range lenses I use are something like 65-75 lp/mm MTF50 (but I haven't seen a way to relate MTF50 to Nyquist limit?)

[ErikKaffehr]

Hi,

You should pay some consideration to MTF 10, low frequencies dominating visual impression of sharpness but high frequencies dominate aliasing. I have seen a paper from Schneider that stated MTF at Nyquist should be below 10% to avoid aliasing. What I see that it takes stopping down to f/16 to avoid aliasing on my P45+, with f/11 aliasing is present at full strength. The P45 has 6.8 micron pixels, so I would assume that 3.4 micron pixels would be safe at f/8 with the lenses I have.

The image below shows MTF of my Planar 80/2.8 on the P45+ without sharpening, and also the MTF of my APS-C lens on the Sony Alpha 77. The P45+ has heavy aliasing at f/8 and the Sony Alpha virtually none at f/8.

AFAIK, MTF 35 - MTF 50 is dominating visual impression. Detail below 35% MTF (or so) is perceived as unsharp. MTF can easily be boosted by sharpening.

So, ideally the system would have high enough resolution that MTF would be near zero at Nyquist. Sharpening can than be applied to make the image visually appealing.




The same image, sharpened:



Best regards
Erik

If I did the arithmetic correctly, the Sony APS-C sensor in my camera are something like 105 lp/mm and the bottom of the range lenses I use are something like 65-75 lp/mm MTF50 (but I haven't seen a way to relate MTF50 to Nyquist limit?)

[Telecaster]

Since when was 35mm sensor 'tiny'?

35mm was introduced as a miniature format. I'd say that's a more accurate description than today's over-inflated "full frame."

-Dave-

[ErikKaffehr]

Hi,

On the film side it is pretty small, on the digital sensor side it is pretty large…

Best regards
Erik

35mm was introduced as a miniature format. I'd say that's a more accurate description than today's over-inflated "full frame."

-Dave-

[Telecaster]

Furthermore, detail is not the quality that purchasers of these photographs have been looking at. They buy the photographs because of their overall emotional impact. Is lack of detail sending some potential buyers away, that I couldn't say. As I alluded to in my first post, the pixel peepers are typically other photographers. A couple of them have even remarked on the lack of details - but they weren't looking at the photograph, only the details; IOW they, too, missed the point. But that's okay. We are all blinded by our assumptions.

When I make prints primarily with other photographers in mind I make them smaller than usual, say 9x12" instead of 15x20", in order to minimize the "nose on paper" phenomenon. Works well!

-Dave-

[Bart_van_der_Wolf]

35mm was introduced as a miniature format.

Not really, it's size was an increased size movie film frame, to achieve better quality but still in an easily portable camera format:

Barnack decided that the 18 x 24 mm (3:4 aspect ratio) standard movie frame was not large enough for good still photo quality with the films of the day and doubled the frame size to 24 x 36 mm (2:3 aspect ratio), with the image horizontal instead of vertical.

Cheers,
Bart

[kers]

Megapixels, smegapixels. Too much hand wringing and analysis, not enough getting out and making photos!
Just forget about it all, go out and shoot what you have and enjoy your photography! :-)

Apart from megapixels being not important for a lot of images that don't need that level of detail and perfection, there is that other phenomena- Dynamic Range.
DR is - i think probably more important - it is about tonality, about a less graphical and a more photographical image.
Printing a 36mp photograph means 30x20cm 600dpi. Combined with a large Dynamic Range, and printed @ 600dpi means you can make a print with rich tonality even if the subject is unsharp.
I love the rich dark tones of the old baryt photographs and in this way we can produce them again in a digital manner.
Looking back i see that the digital image has become more and more photographical with the DR of color negative film. Even the d810 has better tonality than the d800e i find.
So apart from the fact if we need all these megapixels, i like the fact that i can now print a soft image with many tones. ( and choose to make the image more graphical if i want)

[Bart_van_der_Wolf]

Apart from megapixels being not important for a lot of images that don't need that level of detail and perfection, there is that other phenomena- Dynamic Range.
DR is - i think probably more important - it is about tonality, about a less graphical and a more photographical image.
Printing a 36mp photograph means 30x20cm 600dpi. Combined with a large Dynamic Range, and printed @ 600dpi means you can make a print with rich tonality even if the subject is unsharp.
I love the rich dark tones of the old baryt photographs and in this way we can produce them again in a digital manner.
Looking back i see that the digital image has become more and more photographical with the DR of color negative film. Even the d810 has better tonality than the d800e i find.
So apart from the fact if we need all these megapixels, i like the fact that i can now print a soft image with many tones. ( and choose to make the image more graphical if i want)

I'd say they are both important. But don't confuse sensor DR with output DR (it's a common mistake people tend to make). Output DR in print is often not more than 7 stops, because a optical D-max of about 2.10 is already very hard to achieve in print, and D-min (paper white) around 0.10. The effective delta D of 2.0 equals 6.98 stops.

The tonality differences we get from our sensors, after demosaicing, gamma, and tonecurve adjustment, are in principle more than enough for smooth gradients and locally boosted clarity and detail. We then need to output at 8-bit/channel, coming down from 16-b/ch original image data after processing. Having more pixels also helps to make better and more accurate tonality transitions.

Cheers,
Bart

[Telecaster]

Not really, it's size was an increased size movie film frame, to achieve better quality but still in an easily portable camera format:

Well, yes, 35mm was/is larger than the standard motion picture format…but smaller than the popular still picture formats of the time. My dad started using 35mm cameras in the late 1950s only because he really liked the look of Kodachrome, and Kodachrome was unavailable in his preferred 120 format. Even my first camera was a c. 1960 Kodak 127-format Brownie. 4x6cm negatives. The idea of a 24x36mm still image size being large—"full" even!—goes back barely more than a decade. It makes sense, I guess, given that popular still format sizes have consistently decreased over the past century or so. Even so, if photogs in the 1970s–90s ever referred to ~6x7cm as "full frame" compared to 6x4.5cm I musta missed it. 

-Dave-