The Mega Pixel Myth

[smilodon]

I posted this on a private photo forum I belong to and thought it wouldn't hurt to cross post it here in case anyone finds it useful.


When out looking for a new camera or choosing our next smartphone we usually get two numbers shoved at us in an attempt to convince us to buy a specific product. They are price and mega pixels  Price is a simple enough concept and we clearly need to think about the cost of a camera when buying it. But mega pixels is a less clear factor. While shops and advertisers push higher mega pixel counts as being a better thing they don't really explain why that is and certainly don't  mention the fact that high mega pixels can often lead to a poorer image rather than a better one.


First off we need to think about what a pixel is as it relates to a camera. We have pixels on computer monitors and dots on printed images but pixels in a camera is something slightly different. When a camera captures an image and stores it as a picture it actually creates it from millions of tiny dots. Each dot is a certain colour and brightness and when we view these millions of dots on a computer screen or in  a printed image the dots are blended together to create a smooth image. So pixels are the digital dots that make up a photograph. A million pixels is a mega pixel and there you are.


So why is more pixels better than less pixels? A picture is no more than the pixels that it comprises of. There's no additional data that can contribute to the image. So as we enlarge an image on a screen or in print we can't add more information to the picture (Actually we can in software but the technique, called re-sampling, is a whole subject itself and not a substitute for real pixels captured in the camera). Enlarging a picture therefore means enlarging the pixels as they appear on the screen or page. At some arbitrary point the pixels will become large enough that our eye no longer blends them into continuous tones. We begin to see the pixels themselves as a grainy effect. Usually this isn't a good look and the picture basically starts to fall apart. At what point this effect appears depends on the number of pixels in the image, the distance we view the image from and how good our eyesight is. But it's clear a one mega pixel image will start to degrade quicker than a ten mega pixel image. So the more pixels we have the larger we can make our image while keeping it looking nice. We need more pixels to print a picture six feet across than we do to print one on a postcard. And that's about it, there's no real benefit to pixels beyond the increase in size we can render the picture. Anyone who prints no larger than a 8" by 10" picture frame or a web gallery can happily live in the sub five pixel zone. In fact you could probably print a five mega pixel image much larger than that with no ill effects. Having a ten or fifteen mega pixel camera will not really help make your pictures any better and may well make them worse as we'll see in a moment.


However it's worth thinking for a moment about cropping images. More mega pixels can help quite a bit if we crop images down. Assume we have a five mega pixel image that we want to print quite large. That's not a problem as five mega pixels is a lot of dots. But assume we need to cut out a lot of clutter from our picture and centre in on the main subject. So we chop out half the image. We now have a much nicer looking photograph but we have thrown away half our pixels and are now down to a total of two and a half mega pixels. When we try to print the image we have to make the pixels twice as big as we would have done had we not cropped the image. It's quite possible that this new smaller picture will 'break up' and be unusable unless we keep the size quite small. Not good.


So why not just grab as many pixels as possible in case we need to print stuff the size of a barn door....? Here's why. A camera captures an image on it's digital sensor. The sensor is covered in millions of ting photo-cells. Each cell captures a single pixel. I think of these cells as tiny buckets that catch photons of light. Each bucket counts the number of photons of light and the colour of the light and assigns it a numerical value. These millions of values are taken by the software in the camera, PC, printer etc and the image is formed from them. So a five mega pixel camera will require five million little buckets on it's sensor while a ten mega pixel camera will need ten million buckets. The problem is that while pixel counts have been rising the size of the sensor generally hasn't. This requires ever more pixels to be crammed onto the same size sensor. As an example the pixel count of the iPhone has risen several times but the sensor size has not. The problem with this ever growing number of pixels is two fold.


Firstly more pixels creates more heat and heat creates 'noise'. Noise is the peppery grainy effect that can sometime be seem in images. Occassionally if we're going for an old style image or a gritty urban landscape this grain can look effective. However it's usually best to take a 'clean' photo and then add the grain selectively in software. Cameras and software can do great things removing noise from images, but it's always better to start with little or no noise in a photograph rather than trying to remove it later.


The second issue is that the more pixel buckets that are squeezed onto a sensor the smaller they become. Smaller buckets collect less light and so cannot make the subtle distinctions between tones that cameras with bigger buckets can. This leads to flatter less contrasty images that lack richness of colour and tone.


So it's not unlikely that taking pictures with a high pixel camera will lead to an inferior image when compared to one taken on a lower pixel camera. Only if the image is being viewed or printed very large will it potentailly be worth loosing out on tonal range and low noise to retain the additional detail the extra pixels provide. Big megapixel cameras are not necessarily a bad thing and on full size DSLR's it's unlikely to be an issue using the twenty plus mega pixels they provide. These effects are often subtle and normally only really show up on very cheap cameras when used in difficult situations such as very dark, very light or very contrasty conditions. However we can spend a lot of money on camera phones and compacts, prices that approach the price of a full DSLR. There are many very good web sites that give honest appraisals of most modern cameras with detailed analysis of the images they can create. DP Review is one of my favourites and has the nice feature that the reviews have a nice short summary and a list of plus and negative points. So there's no need to read through masses of techo-babble to find out if the camera you're interested in is worth buying.

[Benny]

If you haven't, you should blog that, it's very interesting. I want to be a good photographer but am painfully not, I have neither patience nor understanding to be so, but little nuggets like this keep me interested. Cheers Smilo.

[T-Bag]

Very well written. Gets the point across perfectly. The smaller the sensor the more susceptible to noise it is. There are important contributions to the signal which are temperature dependent (Dark Current), the smaller your total signal the more significant this effect. I've got a lab camera which is 1 mega pixel but has very low dark current because it is cooled to -40C so works incredibly well in low light conditions (ie below what the naked eye can see). Add more pixels to it and I'm sure you'd negate that and just see noise.

On top of that you have quantum efficiency. A good spectrometer will absorb 90% of light hitting it. A scientific grade camera maybe 70% (in the visible range and near IR), an SLR I suspect much less. That means for every 10 photons hitting it more than 3 are just thrown away. You can have as many megapixels as you like if your sensor has a crap quantum efficiency you'll be throwing away most of your image.

There's far more to an image than the megapixels or even the sensor size. Read reviews which cover the actual image quality and you'll learn far more than you will from the resolution. Also rememeber if you put a rubbish lens in front of an amazing sensor you'll be disappointed...the image is only as strong as the weakest link.

[smilodon]

If you haven't, you should blog that, it's very interesting. I want to be a good photographer but am painfully not, I have neither patience nor understanding to be so, but little nuggets like this keep me interested. Cheers Smilo.

You're based down south? We should go on a photo walk one weekend. We can wander about, find things to photograph, chat, get bored and go to the pub.

I've been on a few properly organised Photo Walks/Workshops and they're great for getting to grips with photography. I may be the worlds worst landscape photographer but I went on a Photo Walk a few years back with a Chilterns Group. The leader was really helpful and now I consider myself to have improved to the level of terrible.

T-Bag I had a vague idea about how heat created interference in the sensor but no real solid idea about dark current. Thanks for shining a light on that issue....... see.... see what I just did there....... shine a light? Suit yourselves :sad:

[Dingo]

Perhaps it's just me, but, all I got from that was........


Every pixel tells a story
A pixel is worth more than a thousand words




......and as for T-Bag's comments, well, it just appeared a bit black and white to me!!:D

[DrunkenZombiee]

Good article Smilo but there is another level you have missed here...

Most Modern cameras are Bayer arrays where you have a single pixel representing only a single colour. Worse than that there are 2 time the number of green Pixels to Blue and Red pixels so realistically you are averaging the light intensity over 2 pixels (Some would argue 4 or more as all pixels around that pixel will contribute). In real terms an 8MP sensor is actually half or quarter that as the light is averaged out over these 4 pixels and the pixels above and below. So a pixel in your camera doesnt have an R,G and B value like your computer screen.

A small sensor would look something like this:

RGBGRGBG
GBGRGBGR
BGRGBGRB

There are ways of getting round this. Some very expensive cameras have 3 sensors (one for Red, Green and Blue) and use a prism to split the light into its different colours for these sensors. But these are very complex to make and heavy due to the prism.

Hope this helps someone else.

DZ

[sulky_uk]

i just seen a new camera that we may be getting in the bay........286 mega pixel aircraft mounted camera....wow

[DrunkenZombiee]

The sensor size will be very large though so each "Bucket" will be largeish. Also as its on an aircraft there is no issue with size, weight to a certain extent (walking around with a 50kg camera is a no no) and cooling as power consumption wont be an issue. But its attached to a plane so not really good for portraits or use indoors =P

[T-Bag]

I plus replacement lenses for indoor shooting would probably be hard to come by.