11

Do more megapixels mean sharper images?

Would a 24 MP camera produce a sharper image than an 18 MP one (given the same lens, settings and scene)?

14

Higher megapixels do not add to lens sharpness.

This has been found by many Canon EOS 90D owners. It has 32.5 megapixel APS-C sensor. Its pixel density is the same as 83.2 megapixel full frame camera. For example, Canon has announced a list of recommended lenses for EOS 5DS that is a 50.6 megapixel full frame camera.

Interestingly enough, Canon has not announced a list of recommended lenses for EOS 90D. I think that's because there are none! The APS-C lenses are cheap anyway so none of those can handle the 32.5 megapixel resolution, and the full frame lenses good for 90D would need to handle 83.2 megapixels of resolution to be sharp on 90D in the center of their image circle!

I think you'll find that on an APS-C sized camera, the practical limit of usefulness for megapixels is at somewhere around 20 megapixels, and for full frame cameras, the practical limit of usefulness for megapixels is at somewhere around 50 megapixels.

| improve this answer | |
  • Just for reference, what would be the practical limit for a phone-sized camera lens, considering that they are currently at 48+ megapixels? – user2813274 May 5 at 16:06
  • 4
    @user2813274 Indeed, for example Samsung S20 Ultra has 108 megapixels. Or so they claim. Actually, when shooting in dim light (hint: indoors is dim), they use 3x3 = 9 pixel areas that have the same color mask. So the Samsung S20 Ultra which claims to be a 108 megapixel camera actually is more like a 12 megapixel camera. – juhist May 5 at 16:24
  • 2
    You can just map multiple pixels to the same sensor output. Those extra pixels will be completely redundant, but then it will pass the "moar pixels" test and sell better. The pixels need to match the sensor pixel density. – Nelson May 6 at 6:30
11

Not necessarily. Each lens can only produce a certain amount of detail, which means at some point it doesn't make sense anymore to increase pixel count because the given lens is not good enough for it.

Higher megapixel sensors are also more vulnerable to camera shake and usually perform worse in low lighting conditions (as the individual pixels are smaller and can gather less light).

Other than that, higher resolution means more detail (which I think you refer to as "sharpness"), especially when zooming in.

| improve this answer | |
  • There's a trend to ramp up mp counts to ridiculously high numbers and group 2x2 or 3x3 blocks into "superpixels" for low light. So far I've seen this in phones, but if it works well, it could make its way to cameras. – xiota May 4 at 15:44
  • But does higher resolution not mean higher MP? Sorry, I am a beginner here. – user91798 May 4 at 16:04
  • @KieranBrown it does! Resolution just describes how many pixels there are. Megapixels are a common "unit" for this with one megapixel being 1 million pixels. This means that if your camera has 20mp, there are 20 million individual pixels on the sensor – Jonas May 4 at 17:37
  • 3
    @Jonas Well, the eventual resolution of an image is determined by lens and pixel density and pixel sensitivity and noise; as you say yourself, when the pixels are not the limiting factor throwing more of them at a bad system does not improve anything. – Peter - Reinstate Monica May 5 at 5:21
  • 3
    @Jonas The term "resolution" was around in photography long before pixels and digital imaging existed. That is far from all that it can mean! – Michael C May 6 at 5:26
9

It depends on the lens and the aperture at which it is used. Even a theoretically perfect (diffraction limited) lens can only generate an average of ~16MP at f/11 on a full frame sensor, and it goes down by ~50% for each 1.5x crop factor (7MP for APS, 4MP for 4/3, green wavelength).

This chart shows the maximum resolution theoretically possible at each aperture for blue/green/red wavelengths (green is the most important for most sensors).

enter image description here

https://www2.uned.es/personal/rosuna/resources/photography/Diffraction/Do%20sensors%20outresolve.pdf

But our perception of sharpness has a lot more to do with contrast than it does with actual detail/resolution. And it also varies with the size of the detail; e.g. a larger detail/area requires less contrast to be resolved/sharp. Most resolution charts/measurements are based upon 50% contrast (MTF50)... but that is a little bit arbitrary IMO. DXO uses the human optical response curve rather than a fixed contrast level when determining system resolution (perceived MP's).

| improve this answer | |
  • Interesting chart. It should be noted that the theoretical limit at large apertures is not reached by most lenses because lens errors dominate the resolution. The overall optimum resolution is somewhere in the middle. – Peter - Reinstate Monica May 5 at 5:22
  • 2
    And what the table shows is actually not resolution but the the absolute number of megapixels above which adding more pixels is useless, for each format, aperture and wavelength. And the highlighted numbers at f/11 indicate typical use cases. Very interesting paper. – Peter - Reinstate Monica May 5 at 5:41
  • @Peter-ReinstateMonica, assuming the sensor can resolve to that level the resolution and megapixels are the same thing. – Steven Kersting May 6 at 13:03
2

I have a 20MPx compact camera and an old Nikon that has 5MPx. I have configured the compact one to also only capture at 5MPx because it's optics are not good enough to capture any detail beyond that detail in normal usage.

Optics is responsible for projecting the image on the chip. Once the scale of blurriness is bigger than a single pixel, having more pixels doesn't make the picture any better.

Good test for this is to take some pictures and then zoom them on your computer. If the image is smudged over multiple pixels, you can reduce resolution on your camera to save memory space and time.

Another thing to remember is that 4K monitor, which very few have has little over 8MPx. So larger resolutions are only important if you want to crop the photos, print them on large wallpapers or something like that.

| improve this answer | |
1

I'm new here thus I cannot reply to comments. I'm agreement with all the responses above. I will also add that each output format is has an optimal resolution. If one wishes to print a full, 'uncropped' image, then the RAW file should be converted to JPEG image at the appropriate resolution. And if the image is sharp and clear at that output resolution, then that's all you need. A 24 MP is more than what you need for a 300 dpi 8*10 print (3000 * 2400 pixels), thus when exporting to JPEG, one could choose a lower resolution than 24 mp. Even fewer pixels for a 5*7 or s desktop wallpaper. I agree that the 90D sensor resolution is a bit absurd, although if one can get an 100% sharp image, it opens up opportunities.

| improve this answer | |
0

Do more megapixels get you a sharper image? Only to a point.

TL;DR: This sensor is still not overkill for high-quality low-focal ratio glass.

Dawes' Limit

The maximum resolving detail is defined by the lens and some rules of physics based on the wave nature of light.

The simple formula given by Dawes' Limit. Also Rayleigh Criterion and Angular Resolution are based rules of physics that assume perfect optics. These are based on the size of the objective lens.

E.g. assume a 300mm f/2.8 lens. This has an objective lens diameter of about 107mm or about 4.2 inches.

Dawes' Limit suggests you find the resolving power of the lens given by this formula:

R = 11.6/D

Where R is the resolving power in arc-seconds and D is the diameter in centimeters. 107mm is 10.7cm.

This gives us

R = 11.6/10.7 ... so the value for R is 1.08

The lens (never mind the camera) can resolve just barely more than 1 arc-second -- again, assuming perfect optics (these are rules of physics... not based on optical engineering and manufacturing tolerances. This is what you get if the optics are flawless.)

Image Scale

A 300mm focal length an an APS-C sensor (1.6x crop factor) gets a field of view which measures 4.3° x 2.9°. Converted to arc-seconds, this is 15,480 x 10,440.

The camera sensor resolution is 6960 x 4640. The pixels are 3.2µm. If we divide that out, we get:

Horizontal: 15,480 ÷ 6,960 = 2.22 Vertical: 10,440 ÷ 4,640 = 2.25

This means we're getting just over 2.2 arc-seconds per pixel (that is our image-scale per pixel at this focal length).

The lens can provide a resolving power of 1.08 arc-seconds ... and the camera sensor can record just about 2.2 arc-seconds per pixel. By this measure, the camera is under-sampling.

Nyquist-Shannon sampling theorem

Nyquist-Shannon sampling theorem basically you should sample at double the resolution of the subject. Since the lens is providing 1 arc-second resolution, what you really want is a sensor that can record 2 pixels per arc-second (double what the lens provides). What we really have is one-quarter of that (now we're really under-sampling).

I picked a big lens on purpose. If you do this again with a smaller lens, you wont get 1 arc-second resolution ... for example a 70-200mm f/2.8 lens has roughly 1.6 arc-second resolution.

The main point is... it is still possible to drive up resolution.

Diffraction-Limited photography

The reality is... the lenses aren't perfect and probably don't offer the theoretical maximum resolving power. Also you're not necessary shooting wide-open and other issues of diffraction limits kick in.

This again, deals with the wave-nature of light. As you constrain the aperture size, light "bends" as it passes the edge due to its wave nature and this reduces the resolving power.

You can read about diffraction limits and see some calculators that let you play with them here: https://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

Bottom Line

Congratulations on making it this far. The camera sensor resolution is still not too high when using high quality glass and low focal ratios. If you start stopping down, diffraction limits kick in sooner than most other cameras (at which point the camera is over-sampling).

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy