Understanding the relevance of megapixels
Discussion
I was under the (presumably mistaken) belief that the more mega pixels, the better quality the picture. Taking it to the extreme, a picture made up of 50 megapixels is clearly going to have virtually no resolution whereas one made up of 50,000 is going to be much higher resolution.
So that being the case, given that the Nikon D4 has replaced the D3, why has it got 16mb compared to the D3x at 24.5? (although the D3s has 12.1). Surely they've want to up the resolution from the old model?
Likewise, the much cheaper D800 has an astonishing 36mp and even the D600 has more than the flagship D4 at 24.3mp.
Very confused by this. I appreciate that there's more to a camera than mega pixels, but surely they're pretty fundamental to the final picture quality?
So that being the case, given that the Nikon D4 has replaced the D3, why has it got 16mb compared to the D3x at 24.5? (although the D3s has 12.1). Surely they've want to up the resolution from the old model?
Likewise, the much cheaper D800 has an astonishing 36mp and even the D600 has more than the flagship D4 at 24.3mp.
Very confused by this. I appreciate that there's more to a camera than mega pixels, but surely they're pretty fundamental to the final picture quality?
First up, remember that the size of the sensor is v. important. Some camera phones talk about 12 mp... but the sensor is TINY, so the photos are crap and noisy. Mp is largely a marketing ploy in cheap cameras/phones as people seem to think more mp = better.
Then go figure that my D3 takes as good a photo as my D800 (which has 3 times the pixels). However, if I crop the D800 image - or enlarge to a bigger physical size, I'll have more info to play with.
Then go figure that my D3 takes as good a photo as my D800 (which has 3 times the pixels). However, if I crop the D800 image - or enlarge to a bigger physical size, I'll have more info to play with.
Megapixels determine the resolution of the image. It does not determine the image quality, which is what it is used in marketing as.
Resolution is one part of what determines image quality.
You could have a 50mp image that is absolutely terrible quality and you could have a 5pm image that is superb quality.
There is so much to take into account when looking at image quality that "megapixels" is almost irrelevant.
Take Sigma cameras and the foveon sensor. They would produce a 6pm image that could be compared in quality to the likes of normal DSLRs producing 20mp+ images.
Megapixels is simply a marketing term used to simplify how good a camera is. To the general public, a bigger number equals better product and retailers can latch onto this and command a higher price for cheap tat that has a big number attached to it.
In the real world, for normal use. 10-12mp is about right. I certinately wouldnt consider going higher on anything with a smaller than APS-C sized sensor anyway.
Resolution is one part of what determines image quality.
You could have a 50mp image that is absolutely terrible quality and you could have a 5pm image that is superb quality.
There is so much to take into account when looking at image quality that "megapixels" is almost irrelevant.
Take Sigma cameras and the foveon sensor. They would produce a 6pm image that could be compared in quality to the likes of normal DSLRs producing 20mp+ images.
Megapixels is simply a marketing term used to simplify how good a camera is. To the general public, a bigger number equals better product and retailers can latch onto this and command a higher price for cheap tat that has a big number attached to it.
In the real world, for normal use. 10-12mp is about right. I certinately wouldnt consider going higher on anything with a smaller than APS-C sized sensor anyway.
Edited by MysteryLemon on Tuesday 5th February 14:36
Analogy followed over from other thread.
Imagine a football field. Put out buckets all across it. The buckets need to catch rainwater and then they do something (turn from green to red or vice versa, doesn't matter what). Just imagine it needs FIVE raindrops to trigger each bucket. Each bucket has a circular profile to the sky. Some raindrops will fall between the buckets
Now, put more buckets out......by making each bucket smaller!
You've got more megabuckets. Because buckets are round, you'll catch more raindrops as they pack together well. That's good. But because the buckets are smaller, it takes more time to collect the necessary five raindrops. Some of the raindrops even evaporate (the buckets warm up!!!) before five get there!!!!!
OK - simple analogy. Lots of holes! But you get the picture. When the buckets warm up and evaporate the drops that's like "noise" on a digital sensor. Accepting less drops means more "noise" until the signal gets swamped!!
Of course the sensor size is easy to see now.
imagine catching all that rain in something the size of a handkerchief!!!!
Physics sort of bv$$3rs things up a bit!
Imagine a football field. Put out buckets all across it. The buckets need to catch rainwater and then they do something (turn from green to red or vice versa, doesn't matter what). Just imagine it needs FIVE raindrops to trigger each bucket. Each bucket has a circular profile to the sky. Some raindrops will fall between the buckets
Now, put more buckets out......by making each bucket smaller!
You've got more megabuckets. Because buckets are round, you'll catch more raindrops as they pack together well. That's good. But because the buckets are smaller, it takes more time to collect the necessary five raindrops. Some of the raindrops even evaporate (the buckets warm up!!!) before five get there!!!!!
OK - simple analogy. Lots of holes! But you get the picture. When the buckets warm up and evaporate the drops that's like "noise" on a digital sensor. Accepting less drops means more "noise" until the signal gets swamped!!
Of course the sensor size is easy to see now.
imagine catching all that rain in something the size of a handkerchief!!!!
Physics sort of bv$$3rs things up a bit!
What they all said basically but the case of the D600 vs D800 vs D4 is an interesting one. They're all full frame sensors and you could make an argument to say that the D600 and D800 have better sensors than the D4 in terms of absolute image quality and that in many like for like circumstances would produce better results.
However the D4 has a couple of trump cards up it's sleeve. Firstly it performs quite a bit better at higher iso, almost certainly because it has less mp and therefore each sensor cell is larger or pixel density on the sensor is smaller. Secondly it's smaller file sizes allow it to bang out 10 frames per second while a D800 struggles along on just 4. For many pro shooters, these two things would be of far more use than being able to print enormous images. 16mp should still be enough to print very crisp A1 images.
This was also the case with the D3X vs. D3S. The D3X is not known for it's ability to shoot in the dark. You used that camera if you wanted ultimate IQ in controlled conditions. The D3S was the high iso, machine gun shutter camera that would have suited more people's needs. Things are a bit cloudier now as the D800 and D4 are very different cameras and the D800 actually performs very well in low light but the same differences still exist.
In terms of small compacts and camera phones boasting huge pixel counts, it's unlikely that they would ever actually be able to resolve detail that fine which basically in my head means that their sensors are out performing their lenses which makes the actual mega pixel count an irrelevance.
However the D4 has a couple of trump cards up it's sleeve. Firstly it performs quite a bit better at higher iso, almost certainly because it has less mp and therefore each sensor cell is larger or pixel density on the sensor is smaller. Secondly it's smaller file sizes allow it to bang out 10 frames per second while a D800 struggles along on just 4. For many pro shooters, these two things would be of far more use than being able to print enormous images. 16mp should still be enough to print very crisp A1 images.
This was also the case with the D3X vs. D3S. The D3X is not known for it's ability to shoot in the dark. You used that camera if you wanted ultimate IQ in controlled conditions. The D3S was the high iso, machine gun shutter camera that would have suited more people's needs. Things are a bit cloudier now as the D800 and D4 are very different cameras and the D800 actually performs very well in low light but the same differences still exist.
In terms of small compacts and camera phones boasting huge pixel counts, it's unlikely that they would ever actually be able to resolve detail that fine which basically in my head means that their sensors are out performing their lenses which makes the actual mega pixel count an irrelevance.
Edited by Gad-Westy on Tuesday 5th February 15:24
Gad-Westy said:
However the D4 has a couple of trump cards up it's sleeve. Firstly it performs quite a bit better at higher iso, almost certainly because it has less mp and therefore each sensor cell is larger or pixel density on the sensor is smaller. Secondly it's smaller file sizes allow it to bang out 10 frames per second while a D800 struggles along on just 4. For many pro shooters, these two things would be of far more use than being able to print enormous images. 16mp should still be enough to print very crisp A1 images.
This, for press, sports photographers etc they're not that likely to be making massive prints, but the conditions they're working in are likely to be sub optimal, premiership football grounds in the dark with heavy rain, getting a good shot in low light with enough FPS to catch the action is far more important than filling a billboard.The guys needing massive resolution (high end commercial work, architectural stuff etc) tend to use medium format digital, the resolution blows dSLRs away but they have worse high ISO handling, not as big as issue for this kind of work where generally you can control the lighting.
So bigger pixels will "grab" the light quicker than small ones and therefore speed up the shutter speed?
My take on it is, imagine building a picture ten feet by ten feet using different coloured Lego blocks. Surely the smaller, and therefore the more, blocks the smoother and less "blocky" the picture?
My take on it is, imagine building a picture ten feet by ten feet using different coloured Lego blocks. Surely the smaller, and therefore the more, blocks the smoother and less "blocky" the picture?
Ari said:
So bigger pixels will "grab" the light quicker than small ones and therefore speed up the shutter speed?
My take on it is, imagine building a picture ten feet by ten feet using different coloured Lego blocks. Surely the smaller, and therefore the more, blocks the smoother and less "blocky" the picture?
Yes, that's right. The problem is that the blocks will be sMy take on it is, imagine building a picture ten feet by ten feet using different coloured Lego blocks. Surely the smaller, and therefore the more, blocks the smoother and less "blocky" the picture?
tty colours (almost random colours in some cases).Make sense?
Light is made of photons. The smaller the pixel, the fewer photons hit it.
ATEOTD day you only need so much resolution; the rest just makes for unwieldy file sizes.
I shoot weddings mostly with a 10Mp camera. Horrfying ain't it? But hundreds of thousands of photos later, nobody has said 'I wish your photos had more pixels in them'.
ATEOTD day you only need so much resolution; the rest just makes for unwieldy file sizes.
I shoot weddings mostly with a 10Mp camera. Horrfying ain't it? But hundreds of thousands of photos later, nobody has said 'I wish your photos had more pixels in them'.
Remember, also, that sensors can only cope with one-wavelength-at-a-time-pixels.
Hence, dealing with colour is an interesting game.
Sensors have a pattern of different sensors, typically for red, blue and green.
This makes things tricky when a red photon arrives at a blue sensor, et cetera. Then, the camera has to be too-clever-for-its-own-good in interpolating what it should be seeing. Hence why digital images do funny things and need sharpening somewhere, either in camera or in PP.
(Film used to deal with this quite easily, the different sensitive emulsions were arranged in layers. (A bit like layers in Photoshop.) But, of course, layers solution isn't viable for photosites.
Hence, dealing with colour is an interesting game.
Sensors have a pattern of different sensors, typically for red, blue and green.
This makes things tricky when a red photon arrives at a blue sensor, et cetera. Then, the camera has to be too-clever-for-its-own-good in interpolating what it should be seeing. Hence why digital images do funny things and need sharpening somewhere, either in camera or in PP.
(Film used to deal with this quite easily, the different sensitive emulsions were arranged in layers. (A bit like layers in Photoshop.) But, of course, layers solution isn't viable for photosites.
MysteryLemon said:
Not true. I think 
Isn't the foveon sensor arranged in layers like you suggest? At least that's how I understand it.
foveon is an irrelevance.Isn't the foveon sensor arranged in layers like you suggest? At least that's how I understand it.
Today what governs a sensors resolving/detail power is its megapixels. We are pretty much gathering and making use of all the light that falls upon the sensor , splitting this into more pixels is the easy bit.
What matters at the moment is two things. Quantum efficiency (helps high ISO) and read noise (helps dynamic range).
Which is why the 36mp d800 is actually better in lower light (and has more dynamic range) than the 12mp D700. Its a newer sensor with higher quantum efficiency and lower read noise.
This thing about buckets etc is mostly an old fallacy sorry. Perhaps you have smaller (diameter) buckets but they are just as deep (if not deeper) than they were and the total amount of water gathered is still the same (or more).
Note though that the theoretical resolution (in terms of line pars/mm or picture height) of a camera is WAY less than what you can achieve due to the lens you have to put on the front, and as you get smaller pixels diffraction starts to play more of an hand also limiting what you can achieve.
Also note that P&S pixels are much smaller than SLR pixels and so far actually lead in terms of quantum efficiency!
Now D4 vs D3x and D800. The D4 is a sports / performance camera. We cant at the moment shoot 24 or 36 megapixels, get the data off the sensor, process it and sent it to memory fast enough for a 10-12+ frame a second camera.
RobDickinson said:
This thing about buckets etc is mostly an old fallacy sorry. Perhaps you have smaller (diameter) buckets but they are just as deep (if not deeper) than they were and the total amount of water gathered is still the same (or more).
I'm not looking for a tecchy scrap because you read more than me and will win, but I'd like to point out that a deeper bucket doesn't collect more water, it simply holds more 
Simpo Two said:
I'm not looking for a tecchy scrap because you read more than me and will win, but I'd like to point out that a deeper bucket doesn't collect more water, it simply holds more
The bucket analogy doesnt really cover advanced sensor design in general...But can we say quantum efficiency is how many holes are in the buckets?
Ari said:
Ah, ok I think so. So a huge number of tiny pixels might render smoother edges, but will be less accurate in colour rendition?
Yes.But we're not just talking accuracy of a block of colour. We're also talking about the edges/joins between colours looking rubbish because even though there's lots of small dots, it they're the wrong colour then the edge will look blurry and soft anyway. The camera might also not get the colours right in low light etc. Mega pixles is literally just how many pixles there are. But if each pixel is registering an untrue version of real life, then the photo will still be bad.
The thing some of the other guys are discussing is whether a larger pixel will be more sensitive to light and might produce a more true representation of the light which hit it. This can result in more true to life colours. And where you've got two colours next to each other it can even produce a better defined edge to objects too.
You're better off having 8million pixels which accurately reflect the scene, than having 32million crappy pixles which don't relfect what happened. This is where stuff like what lens you're using comes into play, as well as the size of the sensor and quality of the sensor etc.
And yet the 36mp d800 has a greater colour sensitivity and tonal range than the 12mp d700.
There is a trade off with the bayer filter between colour accuracy and range and light blocking which leads to more noise.
Also consider we have an anti alias filter on these sensors which will blur any sharp lines anyhow and somewhat blend colours between pixels.
Higher MP cameras have by definition weaker AA filters and tend to be affected by moire more. But they still retain more detail even when downsampled to lower resolution (than a nativly lower resolution camera) because this counteracts the AA filter.
IE a 36mp D800 shot downsampled to 12mp will retain more detail than a 12mp d700 shot would.
There is a trade off with the bayer filter between colour accuracy and range and light blocking which leads to more noise.
Also consider we have an anti alias filter on these sensors which will blur any sharp lines anyhow and somewhat blend colours between pixels.
Higher MP cameras have by definition weaker AA filters and tend to be affected by moire more. But they still retain more detail even when downsampled to lower resolution (than a nativly lower resolution camera) because this counteracts the AA filter.
IE a 36mp D800 shot downsampled to 12mp will retain more detail than a 12mp d700 shot would.
Gassing Station | Photography & Video | Top of Page | What's New | My Stuff