Mega megapixels

Bill Pierce

Well-known
Local time
5:00 PM
Joined
Sep 26, 2007
Messages
1,407
A statement that you hear a lot these days from the tech experts is “You really don’t need that many megapixels.” I’m not so sure they are correct, at least in one department, big prints. All other things being equal (and they rarely are), the bigger you make a print, the more pixels you need to hold fine detail.

Here’s a rough guide. Multiply the height and the width of the print you want by the dpi you want. Say I want an image size of 14x21 inches (2x3 i.e. 35mm proportions on 17x22 inch paper with a 1/2 inch border) at 360 dpi on an Epson printer. That’s hardly a mural. But that’s 5040 pixels by 7660 pixels. Multiply the width by the height and you get 38,102,400 pixels or, ballpark 38 megapixels to make that print without upsizing.

It goes without saying that if you are going to utilize the quality of that big count, you better have good lenses, a high shutter speed or good tripod and the blessing of the gods. And to keep those pixels big and noise free a large sensor is an advantage. Admittedly, this is a gross simplification, and there is much to be said for lower pixel counts, especially in high ISO performance and dynamic range (and $$$). But if you want to challenge what folks do with bigger formats in both the digital and film world, I think you are going to need those pixels. With apologies to Edward Weston, who thought 8x10 prints were enough, what do you thing about pixel count?
 
Hi Bill,
I think its important to test the particular printer to see what the minimum dpi it needs to make an acceptable print. I have found that in practice anywhere from 200-360ppi will yield visually similar results on a 24 by 36 inch epson / HP print.
 
Bill Pierce;2389363I’m not so sure they are correct said:
With apologies to Edward Weston, who thought 8x10 prints were enough, what do you thing about pixel count?
And useful heavy cropping as well.

Except "no one prints anymore". For the average user it is something not usually thought of.
Screens have become the main digital medium for output. Increasing in resolution, and 4K is roughly around 10MP. Probably there will be some action in this side.

Writing this from a ~5" screen with a 441ppi density. Still around 2MP "full HD".

Nowadays we'd be better improving other specifications of the sensors. Minimum resolution available seems to be around the 8-12MP in smaller formats... 8x10 without a sweat.
And I've heard saying that for bigger print sizes and accounting for viewing distance, some of the resolution constraints can be relaxed a bit.

For the ones that care about large prints and needing sheer pixelage there are some options available. The nikon D800/810.

While I now have a provision for large prints (Fuji 6x9) I haven't been able to exploit this quality yet.

I am still quite young and try to see things from every vantage point available. For the great majority... MP don't matter.
 
Hi Bill,
I think its important to test the particular printer to see what the minimum dpi it needs to make an acceptable print. I have found that in practice anywhere from 200-360ppi will yield visually similar results on a 24 by 36 inch epson / HP print.

My experience pretty much parallels yours until I go to an exhibit of large prints and see folks with their noses resting against the surface of the print. But, since I don't care to have nose prints on the surface of my pictures, I still pretty much agree with you.
 
drum scan an 10x8 and they can lean their noses on it all day long if it thrills them.

But that is mega megapickels. :)

It really depends on what you intend to do with the file.

A 7x5 from my old Pentax ist DL2 can be gorgeous.

Not so much at 10x8.

Downright terrible at 17x14.

But I really don't like printing everything at 24x30, sometimes 5x7 is what the doctor ordered.
 
Depends on how far back you're standing from the print I suppose, say for billboards, viewed from a moving car, I'm sure 10 dpi is fine. If it's a in a magazine, viewed from a foot away, you obviously need a lot better.

For professionals, doing this for a living, or for amateurs/artists selling prints, then they need the resolution their customers require, but for the rest of us, 6MP will give a more than acceptable 10x8" print, bigger, if viewed from reasonable distances.

High MP counts were created to sell cameras, not create better cameras or photos. If you need the MP count, then that's happy coincidence, otherwise, if the camera is meeting your needs, why change?
 
I am interested in buying a Nikon D800 or D810, I suppose for the reason you stated Bill, I have made the same calculations as you, and I realize that an awful lot of pixels are required for big prints. I do have some images I would like to print very large - but to be honest, not that many, and so I know that this particular argument is a bit of an excuse in my case .....it is the same sort of excuse I use to convince myself I need to buy any new equipment :^) . But I know I will probably upgrade none the less even because I still find that when I shoot some images, landscapes in particular, when I zoom in details like leaves on trees in the distance they are just not there no matter how careful I am with technique and I would like them to be.

The one thing that worries me slightly about going to larger sensor sizes is not the quality of lenses required when shooting a 36 megapixel sensor. I have bought some of the very best Nikon glass - even though some of it is from prior generations including some manual focus lenses that in their day were top pro lenses, it is still excellent and in my view quite up to the job.

What worries me more is technique. My "camera guy" tells me that to get the best out of those megapixels it is absolutely necessary to really take great care about shutter speeds and shooting technique - or to use a very good tripod. Always. Otherwise the image quality that theoretically is possible, is just not there in reality.

I do not wish to hijack this thread but may I ask about shorter term alternatives such as various image enlargement software packages. Has anyone tried any of these and if so, how do they work? examples I have heard of include Smilla enlarger, Alien Skin Blowup, Perfect Resize etc. Of course I image detail that was not captured because of limits on sensor size cannot be put back into an image by that software but otherwise does it work well enough?
 
This is all I have to say:

rff-43.jpg


(Sorry; I just yesterday made that little addition to the Super, and had to share it somewhere!)
 
I think that Bill's arithmetic is applicable only to enlargements of photographs of test charts and even then only under unusually close inspection. From a practical point of view, the situation is nowhere near so dire, if one considers a more practical situation of a general photograph together with the properties of the human eye. Another simple calculation can establish a more realistic and more favorable result.

The typical human eye can resolve objects that subtend about 1 minute of arc ( 1/60 degree). This visual acuity exists only for the fovea of the eye which is the most sensitive portion of the retina and also only exists for objects which are relatively bright and which have high contrast relative to their surroundings. The eye does not have such high resolution for objects having lower contrast, even in its foveal portion.

If one views a photo from a distance d, and if x is the diameter of the dot, and if each dot corresponds to one pixel, then these two values must satisfy the equation

x / d = tan( 1/60 )

where the angle is 1/60 degree. If we take x = 1/360 inch, the dot pitch on an Epson photo-printer, then the equation yields

d = 1 / (360 tan( 1/60 ))

and the viewing distance will be about

d = 9.55 inch

This value will be the MAXIMUM distance from which one may discern high contrast dots (pixels). Note that this is about the usual viewing distance for an 8x10 enlargement (about A4). Thus the megapixel requirement becomes
9.55 x 360 = 3 438 pixels for a 3:2 sensor size of 3438 x 5157 = 17.7 Mpixel.

At larger distances than this, the typical eye will be unable to resolve the dots (pixels). A big enlargement has the same value of dots/inch, and it is typically viewed at a greater distance (about 1 diagonal), so a typical viewer will not be able to resolve the dots or the pixels from the typical viewing distance... NO MATTER HOW BIG THE ENLARGEMENT IS!

Note especially that this treatment assumes that the object in the photograph has high contrast, for example a test chart with alternating black and white bars. Most photographs have few if any regions of such high contrast at the level of pixels or dots. Instead, they can have as little as 10 times as much resolution. In that extreme case, one could inspect an enlargement at d = 0.1 inch and barely see the dots ... and doing that would require a magnifying glass, even for near-sighted viewers!

For a specific comparison, consider a factor of 2 reduction in the resolution requirement due to practical contast. In that case, the megapixel requirement would be 9.55 x 180 = 1719 pixels and for a 3:2 sensor, the size would be 1719 x 2578 = 4.4 Mpixel. This last result reminds one of the fact that digital sensors began to gain widespread acceptance when their size reached the 6 - 8 Mpixel range. Apparently, the acceptance at this Mpixel level was no accident. And it's no accident, either, that good quality printers all have around 300-360 dpi.

--- Mike
 
My experience pretty much parallels yours until I go to an exhibit of large prints and see folks with their noses resting against the surface of the print.

Agreed... people look at photos much closer than people give them credit here.
 
Agreed... people look at photos much closer than people give them credit here.

That seems to be true when there are small details within the photograph that people are trying to identify, and it doesn't seem to matter if it is 8x10 or 24x30. For example, a landscape picture with a car in it, or a house, or some other distinct but unidentifiable structure like some rocks.

I've even seen people do this with paintings, leaning over a rope to the point of being off balance if necessary.

Face it folks. Pixel peeping is a very popular activity and it doesn't stop with a computer screen. :)
 
This is all I have to say:

rff-43.jpg

:D Pretty much sums up indeed!

Nothing beats darkroom enlarging, especially LF if you have the right equipment.

I do love to drumscan my 6x7 MF and 5x7" LF film, gives me rich non interpolated full color information (for each pixel) digital images over 10 000 px a side to work on large digital prints when I need to. That being said the analog film never has the artificial edge sharpness of the digital sensor, but it has the tonal richness, color depth and resolved actual detail - if you're a detail peeper, drum scanning MF and LF film gives you very rewarding results that is very hard to mach (LF still being unbeaten) with today's digital cameras unless you can buy $50 000+ real digital dedicated MF cameras and MF backs, not counting the high-end lenses they require. While you can drumscan for $30 or less.

IMHO of course,
Margus
 
Megapickles? I love that one!

Megapickles? I love that one!

The point of very little more return has been mentioned, and that is 6-8 megapickles. What hasn't been mentioned in the simple sense, is that it is assumed that megapickles would behave by the square/area ruling like decibels, but in fact they can't because of noise.

Because of noise, the greater numbered megapickles sensors may more inefficient, and there are a number of older cameras that are just plain kick ass at 6-8 megapickles (Canon 30D). Therefore, the major benefit of more megapickles is that having a greater number out weighs noise and so doing makes for achieving higher ISO.

Most decent lenses out perform the sensors on even the latest expensive cameras.

Another issue is how the noise is presented. If you want 'pretty' noise, by a Leica Monochrome.

Elsewhere it has been mentioned that the new digital cameras are now evolving 'laterally' with more features and at basically the same performance level.

It has also been mentioned that printing is no longer a goal with digital photography. I agree, but if the large digital print is your quest, stick with larger format film and a drum scanner as previously stated.
 
Superb post from a relatively infrequent poster. Digging back into the archive to see what else you have to say, Mike!

I think that Bill's arithmetic is applicable only to enlargements of photographs of test charts and even then only under unusually close inspection. From a practical point of view, the situation is nowhere near so dire, if one considers a more practical situation of a general photograph together with the properties of the human eye. Another simple calculation can establish a more realistic and more favorable result.

The typical human eye can resolve objects that subtend about 1 minute of arc ( 1/60 degree). This visual acuity exists only for the fovea of the eye which is the most sensitive portion of the retina and also only exists for objects which are relatively bright and which have high contrast relative to their surroundings. The eye does not have such high resolution for objects having lower contrast, even in its foveal portion.

If one views a photo from a distance d, and if x is the diameter of the dot, and if each dot corresponds to one pixel, then these two values must satisfy the equation

x / d = tan( 1/60 )

where the angle is 1/60 degree. If we take x = 1/360 inch, the dot pitch on an Epson photo-printer, then the equation yields

d = 1 / (360 tan( 1/60 ))

and the viewing distance will be about

d = 9.55 inch

This value will be the MAXIMUM distance from which one may discern high contrast dots (pixels). Note that this is about the usual viewing distance for an 8x10 enlargement (about A4). Thus the megapixel requirement becomes
9.55 x 360 = 3 438 pixels for a 3:2 sensor size of 3438 x 5157 = 17.7 Mpixel.

At larger distances than this, the typical eye will be unable to resolve the dots (pixels). A big enlargement has the same value of dots/inch, and it is typically viewed at a greater distance (about 1 diagonal), so a typical viewer will not be able to resolve the dots or the pixels from the typical viewing distance... NO MATTER HOW BIG THE ENLARGEMENT IS!

Note especially that this treatment assumes that the object in the photograph has high contrast, for example a test chart with alternating black and white bars. Most photographs have few if any regions of such high contrast at the level of pixels or dots. Instead, they can have as little as 10 times as much resolution. In that extreme case, one could inspect an enlargement at d = 0.1 inch and barely see the dots ... and doing that would require a magnifying glass, even for near-sighted viewers!

For a specific comparison, consider a factor of 2 reduction in the resolution requirement due to practical contast. In that case, the megapixel requirement would be 9.55 x 180 = 1719 pixels and for a 3:2 sensor, the size would be 1719 x 2578 = 4.4 Mpixel. This last result reminds one of the fact that digital sensors began to gain widespread acceptance when their size reached the 6 - 8 Mpixel range. Apparently, the acceptance at this Mpixel level was no accident. And it's no accident, either, that good quality printers all have around 300-360 dpi.

--- Mike
 
Thanks, semilog. I hope others also found the post interesting.

There's no size sensor that won't show pixels at great-enough enlargement and/or very close inspection. For example, it turns out that a 3:2 FF sensor with pixels the size of a hydrogen atom would have about 86 PetaPixels (86 thousand-million-million pixels), and if an enlargement were made equal to the size of the solar system, then the pixels would be about 18km in diameter. There's no doubt but that some folks would still complain about that sensor!

Superb post from a relatively infrequent poster. Digging back into the archive to see what else you have to say, Mike!

--- Mike
 
Not an expert on cameras, DPI and prints here, but photog who likes different kind of photography.

I have absolutely no problem with 12MP files to be printed at 8x10, which is most common "large" prints for me and people who have prints from me.
I wasn't in trouble with my 16MP picture printed in large poster size once.
Some of my 12 and 16MP cameras pictures were published in printed media.
No problems. It would be no problem even with 8MP files.
The problem for me as active photographer is in large files. Eating HDD space and slowing down computer.
Even if I would get 24MP camera, I'll switch it to 12MP.
 
I made a nice 11x14 from a D40. Remember those at 6 MP. Granted it was on a laser printer, but was a nice as any print I made from film with Leica cameras and enlargers.

I think 12 to 18 is the sweet spot for most users.
I did some low light stuff that I never normally do, but 1600 iso , D700, and 11x17 print looked great. Some NR.

All that said, 36 MP is really nice to work with and is much like my RB67 or even better.
 
Superb post from a relatively infrequent poster. Digging back into the archive to see what else you have to say, Mike!
I think that Bill's arithmetic is applicable only to enlargements of photographs of test charts and even then only under unusually close inspection. From a practical point of view, the situation is nowhere near so dire [...]
I agree, and my more, ahem, agricultural method of estimating broadly agrees with Mike's numbers. Just working with files and prints and observing my results, I've generally figured that 3 megapixels is enough to get a good-quality A4 sized print (note "good" - not surperb, stunning, mind-blowing etc.; but good). I've also figured that 2MP is enough for an adequate-enough (for most purposes) print again at A4 size. So, using nice, sane, sensible ISO paper sizes, I figure I can get good (adequate) prints with the following megapixelage:

3MP (2MP) A4 (8.3"x11.7") roughly letter size
6MP (4MP) A3 (11.7"x16.5")
8MP (6MP) B3 (13.9"x19.7") roughly 'A3+' 13"x19"
12MP (8MP) A2 (16.5"x23.4")
24MP (16MP) A1 (23.4"x33.1")
48MP (32MP) A0 (33.1"x46.8")

[note that the above is just my rough basis for guessing; precision isn't intended or asserted]

I've included the A3+ reference since that's terminology used for a lot of printers and happens to be the largest size I can print at home (and therefore the largest I usually print).

The above accords with actual photos on my wall. I have a shot of a rainbow lorikeet I took with my Canon 300D printed at exactly A3 and mounted where "nose against the glass" viewing is easy and convenient. Many people have been amazed that it's from a camera with "only" 6 megapixels. I have a lot of shots taken with my Canon 30D printed at A3+ (13"x19") which show plenty of detail. My Canon 5D provides more than enough megapixels for me to print larger than I actually can at home to a very good standard of detail.

...Mike (a different Mike)

P.S. I'll also note, regarding printing, that my view has always been that if it hasn't been printed (even just a 4x6) then it isn't a photograph.
 
Last edited:
Yes you do need mega pixels for mega prints :) Here is a real world example. I just printed 29 black/white prints for a show at The Victoria College in Victoria, Texas that debuts Oct 10. All of the prints except one are 24x36 that I printed in house using digital workflow and a Canon ipf6100.
The majority of the images are pure digital - Leica M8, M9 and one image from an Espon RD1. Lenses are all Leica. About ten to twelve are scanned black and white negatives using Leica M2, Leica M3 and CL, again with Leica glass.
The all-digital images look great at 24x36 - no surprise, the digital Leicas have a lot of pixels. But I was pretty impressed with the scanned black and white images, a few from the late 70s. Yes there is Tri-X grain with images that large but even closer than they should be viewed, the images hold up well. Granted you are dealing with scanned images at 3000dpi but again, they looked pretty good.
I was in Paris last year (at the Maison Europeene de La Photographie) and saw a Joel Meyerowitz retrospective with huge color prints, at least 36x48, from 35mm slides. The prints were spectacular! I don't know if it was digital or darkroom but the colors and size were memorable. OC Garza
 
Back
Top Bottom