Why do pixels matter

An individual pixel does not produce "noise". It produces an electrical output. The smaller the pixel size the more likely that it will misread the photons falling on it and produce a different output than the colour and intensity of light demanded.

A smaller pixel will also accept less light before it runs out of dynamic range, like a small bucket filling up faster than a bigger bucket. Combine lots of pixels misreading the signal and you end up with noise.

In simple terms, there are two major sources of noise : electronic 'read' noise and photon shot noise. Individual pixels do produce some degree of electronic noise. But, except at high ISOs and in the deep shadows of your image, this is sufficiently low to have little impact.

Most photography is dominated by photon shot noise. This isn't a question of small pixels 'misreading' the amount of light falling on them small pixels typically have higher quantum efficiency: they're better at producing photoelectrons in response to individual photons of light. The problem instead is that light itself is noisy.

Just like raindrops, it falls randomly on your sensor. The more you measure, the more this randomness averages out signal grows faster than noise, such that your signal-to-noise ratio improved. So small pixels receive less light during any given exposure and can't average-out this randomness so well they have worse s-n ratio. What seems to be omitted from the article is why a difference does emerge at very high ISOs. Averaging out noise will naturally work up to a certain level. Once your individual sample size in this case dictated by the pitch of each sensel and very low photon flow averaging several unreliable sites becomes less effective than one which has a much greater likliehood or recording a true value.

Hence an advantage- but only at very low signal levels. The why it makes a difference at very high ISOs is touched-on in more detail in this article. So long as the two systems have comparable efficiencies and fill-factor, then photon shot noise averages out pretty nicely. In the simplest terms: smaller pixels tend to have less read noise than larger ones, but not always lower in perfect proportion to how much smaller they are.

These tiny differences can become apparent when you apply massive amounts of amplification. Two points. Are you sure that the noise produced by a small pixel is equal to the noise generated by a larger pixel. Noise is a semiconductor issue and I would assume is based upon the size of the device you are dealing with and in this case it is an optical receptor.

Second is these kinds of articles always mention large photo formats when discussing the need for more pixels. I only print 8.

However, where I do need pixels is when cropping or zooming in in post as one can rarely get a full frame of the object you are photographing. So for me more is better. In most photographic situations, photon shot noise plays more of a role than electronic noise from the sensor modern sensors have spectacularly low read noise levels. The take-home point from the article should be that there's essentially no down-side to having more pixels.

Personally I take a lot of bird photos and like you say they take up a very small portion of the frame so cropping is always required. I'm using a D and never had a noise issue that I could tell. Furthermore for the general public they wouldn't know a noise problem no matter how much it affected the photo. What also needs to be mentioned is at the other end, when a consumer camera with a small sensor has 24 mp.

They will never do the job of 12mp on a full frame. Couldn't care less about noise and resolution. Fact is that my old Canon 5d Classic have much better colors and micro contrast than any of those new cameras. Don't know if it has to do with larger pixels or a more selective CFA, but the new cameras simply can't match the output fom my old Canon 5d.

From what I've read resolution wise film can't compare to high res digitals. Film has limitations. Richard Butler: I don't normally print. This question is hypothetical, just to better understand the implication of this article. Let's say I normally use a 24 MP camera. It has pixels across. Someone tells me that 'a printer' does dpi and therefore I can print 30 inches wide.

Good so far. Their sensor size is the same though. Is this article saying that when printed 30" wide, photos exposed the same way from those three cameras will look the same?

Assuming 30" is the maximum print size for a 24MP sensor, then the print from the 12MP sensor should show a lack of details. But you should not see something different from the 42MP sensor. But the 42MP allows you to print at an even bigger size, or crop To build on what Karroly says: they'll look very similar in terms of noise and tonal quality. The 42MP camera may well look more detailed than the 24MP one, even though you've effectively downsized it.

The 12MP one may look a bit mushy and un-detailed. Karroly: As long as you look at the whole picture, you can print any size with 16Mpx or more, you will see no difference between 16Mpx and Mpx. Only if you crop move towards the picture or do pixelpeeping and look only at parts of the picture, you will recognize differences. Richard says: to build on what Karroly says: they'll look very similar in terms of noise and tonal quality.

The 42MP camera may well look more detailed than the 24MP one, even though you've effectively downsized it If the printer prints only dpi as suggested, how can the extra pixels show more details? Because the printer can't. I wonder how often anybody sees the results of the different signal to noise ratios of pixels? When I view a photo I'm looking at composition, color, subject, and the impression that it makes on me.

Is some edge smooth or not? I don't know because I don't know what the real object has. Same with color. What is the "real" color in today's world and does it really make a difference to anyone other than someone whose product has to be properly depicted color-wise like clothing.

To me most of this is sales hype by the camera companies. This is just simple geometry : given the same print size, the image coming from the 16MP sensor has bigger pixels than the image coming from the MP sensor Sorry, but you are wrong: I said as long as you look at the whole picuture! This means: the viewing distance increase with picture size, that's why 16MPx is always suffizient. Otherwise you need better eyes. If you have eagle eyes, you need more pixels, but only then. You just do not use the same "test setup" than me.

That is why we do not agree on the conclusion. But it does not matter. This discussion is over for me. Thanks, but how did you resize the image of the Nikon , pixelbinding? But then: how did you get higher resolution? We should remember: if you want to see the whole picture no pixelpeeping , about 16MPx equals the resolving power of our eyes. Thats why most professional cams have around Mpx.

Only for special tasks like aerial imaging, more resolution maybe needed. I've seen lots of 40x60cm prints with 24Mp up to Mp cams, but sorry, usually I can't see the difference. In a good light, images from this are quite publishable, both on the web and in magazines. Finally, I'd recommend a boot abut low pixel count camera that has many illustration with older cameras under 10Megapixels:.

When I imaged with a D5 MK ll large prints were no problem. Between these two fantastic software products one should no longer worry about sensor sizes, especially starting at I typically toss images that are even slightly soft. To make a fair compare I think the D image should be both scaled and softened to match the noise level of a7S as close as possible, and only then it would be possible to see if you end up with more or less visible details.

Currently we are comparing bananas and oranges, and what would be the best 'as is' depends on the use case. Well I think the Raw file links should work, if you want to try your own workflow and see what you end up with. Richard Butler, should perhaps have been more precise But sure each of us could put the raws through some conversion, but my point was that it should have been done before the conclusion could be made fairly.

The OPTIMAL for a fair compare would have been if some manufacturer have made two cameras at almost the same time with same sensor size and technology but different resolution.

And not cameras two years apart from different manufacturers. The images are presented with the same treatment, but you want to change that? Michiel, simply to make them comparable From the images shown we can not say that one 12MP images is better than the other. Of course rather than soften one, we could also sharpen the other thus making the noise more clear, to see if this also brings comparable details. If the claim is that we can scale an image down and that the resulting image is just as good, then we would need to make them comparable Of course if it had shown both less noise and more detaild it would be a clear case, but as it does not, it is a matter of what the image should be used for, if one is better than the other.

Is the noise more or less important than the details,. What is best depends on the situation, but we can not compare them and claim that one can replace the other.

IF we should be able to replace, then we would need to see what happens if we also 'unsharpen' or 'soften' the image, so both look similar for the coloured surfaces.

And on THAT image we could compare the details e. And only if at least as good, and not smeared away, it could replace what the Sony gave. Instead of "softened" do you mean use noise reduction so that they would have an equivalent level of noise, and then compare the level of detail? If you are using noise reduction then it's even better to have more pixels as any decent algorithm prime and the like can take advantage of them in short - you can generate a more precise model of the noise you want to remove.

For simpler implementations it would just level the field scaling down is a very simple form of noise reduction. Buy, there is no substitute for your own experience so try playing with files kindly provided by the rest of DPR staff. But that brings us back to square one: The claim comes with a series of JPGs neither proving nor disproving the claim It would be nice if that fiddling were done before presenting such a in my book monumental claim. In my world it would be big news and against what I remember to have learned in physics at the university What we have here are also two years and different manufacturers separating things, so that make the conclusion even more dubious as something that could be generalized as done in part two:.

At best we could claim: "a new camera's images downsampled does as well as a two year older with a quarter the number of pixels. All of this pixel-peeping and obsession with 'sharpness' and technology is of no avail unless you you can compose and use light properly. There's no point in blowing up a lousy, poorly composed subject, poorly lit, to A3 size even if you do have the most fantastic sensor and the sharpest of lenses and the best focusing technology available.

The reality is that 'leading edge' technology is for 'leading edge' circumstances. There are objective and subjective measures. Objective ones mathematics, best known being PSNR, but also others indicate what was stated in the article. Yiu disagree. Then there are subjective ones which are - subjective and your opinion is as valid as any other.

Some like smooth, others are OK with noise as long as there is detail and others want "grain" to have character. You can't say one group is right. Have a nice day. But if we start out measuring on very different things, any conclusion drawn could easily become dubious.

If we measure the acceleration and the pull-power of a sports car and a lorry, each is likely to win in one. But to conclude that either could replace the other would make little sense.

So IF we assume these differences in what we actually compare can be ignored and IF we assume that the benefit outweighs the deficit in practical usage THEN and only then the claim holds. But the big question is if these assumptions are reasonable? AntonJA, I don't see how this is relevant to my comment? The same logic could suggest we should stick to good old technology: Brushes and paint I'm just saying that they got a very interesting claim here, but I would like to see something backing it.

Quite futile as we don't have a choice anyway except the FF cameras. For FF, it is more of a resolution dilemma than pixel size.

A Sony A7s with 12MP will resolve roughly half that. What a higher pixel count gains you is "over-sampling" so you can do away with the AA filter if you plan on down-sampling.

The down-sampling is equivalent to low-pass filtering, which will eliminate most aliasing problems. A long time ago, I was messing around with down-sample up-sample comparisons, back when 12Mp was considered high resolution.

Then I took the original with the same JPG quality settings and generate another output. I posted them here and asked if anyone could tell the difference, and nobody could. In fact guesses were usually backwards of expectations. I could overlay the two images in Gimp and subtract them, and there were very few differences. The differences being A few things I learned from this: 1. The process of exporting a JPG is based on preserving detail at the expense of accuracy.

This indicates they had about the same detail. Down-sampling reduces noise. This may only be true with conventional Bayer arrays and basic processing methods.

It nags me that I may be throwing out important data that I just don't use today, that I may want tomorrow. I think that especially point 3. Noise also reduces the sharpness of edges between bright and dark areas in the picture and can make it look unsharp and a little flat. Sometimes noise can only be seen in shadowed areas, where exposure levels are low. It is common for this shadow noise to show a pattern of coloured dots.

Our regular camera tests consistently show that cameras with smaller sensors fail to achieve this in bright conditions. In contrast, DSLR sensors with larger photosites can usually record the full dynamic range in the subject although you may need to shoot raw files in order to extract all the highlight and shadow details in brilliant sunshine.

Note the lack of detail in the bright areas on the left wall of the house and also under the verandah near the front door. Contrast these areas with a similar shot taken with a DSLR camera, which can capture a full range of tones in both brightly-lit and shadowed areas because of its larger photosites. Five megapixels is probably the limit of resolving power for most point-and-shoot camera lenses. Beyond a certain point, diffraction will begin to reduce the resolving power of the lens-plus-sensor system, as we have discovered from Imatest tests on many 8- and megapixel digicams.

In many digicams, the image processor automatically sharpens the image by default. Subject lighting will also play a role in image quality, especially with small-sensor digicams. In dim conditions, photographers are forced to increase ISO speeds. And really, the amount of pixels should be an indicator of optimal sensor size it is NOT though. Begin typing your search term above and press enter to search.

Press ESC to cancel. Skip to content Home Arts Why do pixels matter when it comes to images? Ben Davis March 16, Heck, even a modest 6 megapixel camera will result in some very good pictures.

What you want to do though is figure out how you will be using your camera. If it is going to be a small camera and you want to be able to zoom into an image without having a zoom lens, then more megapixels may be beneficial. Just be sure that the camera will work for the lighting and style of pictures you take.

After all, if you are shooting indoors and in low light, lower might be better. Megapixels is just one factor in a camera too. There is the quality of the lens, the technology behind the sensor, the ability to focus quickly, use a flash, etc. All of these play into how the camera works so always compare more than just megapixels. Your email address will not be published.

Notify me of follow-up comments by email. Notify me of new posts by email. What Does Megapixels Mean? Leave a Reply Cancel reply Your email address will not be published. This website uses cookies to improve your experience.