I have a question about image quality. Does noise in an image is depends upon "Megapixels" or on "ISO"?
Noise originates due to a number of factors, see:
Increasing the number of megapixels keeping everything else constant (sensor size, technology etc.) will increase noise per pixel, but also has the effect of making the noise finer grained which is less objectionable.
ISO does not by itself increase noise, only if you combine increasing ISO with decreasing shutter time / closing aperture.
It's [probably] worth repeating this again here...
Increasing ISO whilst keeping shutter-speed/aperture constant does not increase noise:
Here is an example, as the ISO 100 shot was underexposed, raising the ISO to 1600 yielded a much less noisy result!
When it comes to ISO, it's a misnomer that increasing the ISO increases noise. It's doing nothing of the sort. The reason high ISO produces noisy images is entirely due to the fact that for a higher ISO, you have to cut the amount of light (signal) hitting the sensor, ultimately increasing noise ratio before ISO amplification even takes place.
The short answer is both, but it requires some explanation.
Megapixels. Yes, but only if we're talking about 12MP vs 24MP for a given sensor size. The issue is not the absolute number of megapixels per se, but the size of each individual pixel. Buckets in the rain is the common analogy used. Essentially, if you have two buckets out in the rain, the bigger one with twice the area as the other, the bigger one will collect twice as much water (or photons).
This is all about signal-to-noise ratio.
Photon hits photodiode --> photodiode emits electron --> electrons are stored in a cell --> analogue charge converted to a digital signal by the analog-to-digital unit (ADU). Roughly, anyway.
The level of amplification that takes places at the ADU stage is determined by the ISO level. The issue is that along the way noise occurs. By noise we mean unwanted signal -- the level of charge stored in the well does not 100% reflect the light reading in a perfect world. There are various reasons for this (some of which are natural and cannot be avoided -- details [here]). At higher ISO levels the signal needs to be amplified even more increasing the impact of this unwanted noise as a percentage of the total signal.
The following numbers are arbitrary, but let's just say at ISO 100 you had 4 noise units out of a signal of 12.5 units collected and that at ISO 400 you had 4 units out of 10 units collected. Note...
- the noise levels before amplification are independent of the ISO selected).
- at ISO 400 1/4 as much light is collected (ISO 400 is two stops above ISO 100). Double the ISO, half the amount of light collected.
The amplification process needs to take the signal to 100.
- ISO 100 (x2): 8/100 = 8% noise.
- ISO 400 (x8): 32/100 = 32% noise.
So an increase in ISO leads to more noise. If our individual pixels had been twice as large so that no amplification was needed (again, this is just to explain the concept) then the following would have been true.
- ISO 100: 4 units of noise for 100 units of total signal (4%)
- ISO 400: 4 units of noise for 25 units of total signal (16%)
... and noise would have been reduced.
Both of those factors can have an impact on noise. Of the two, the effect of ISO is the most immediately visible. All digital cameras exhibit image noise that rises as ISO increases. Depending on the camera, you may see little or no noise at ISO 100, for instance, and proportionally more as you approach the maximum ISO setting for your camera (typically 1600 to 6400 or more). There are plenty of examples of this for just about every camera on the market (reviews for each specific camera will often show examples).
Image size (megapixels) doesn't directly impact noise, but it indirectly impacts noise when you consider two sensors of the same size. In this case, all things being equal, the larger-megapixel sensor will show more noise as a side-effect of trying to "squeeze" more pixels onto a sensor of similar size. Although this can be shown in certain cases within a camera line, this factor can be difficult to pin down because cameras also often have upgraded sensor or processing technology that manages noise more effectively.
There are some great answers on this site already regarding specific aspects of noise. If you search for "noise", you'll find all sorts of great reading.
The primary cause of noise in an image is due to Photon Shot Noise as a result of not collecting enough light... This noise is due to the randomness of light/photons. When very little light is collected it is more likely that sensels (pixels) will receive inadequate or even no light. Think of this as being exposed to rain... a short exposure to a light shower and you won't be entirely soaked/covered.
ISO simply allows you to collect less light (smaller aperture/faster SS).
More MP's (sensels/area) simply divide the light that is collected farther (more affected by the randomness). But if the sensor sizes are the same and the images are displayed at the same physical size, then the images will have the same light per area and thus the same noise characteristics, regardless of the MP count.
Some nice answers already. My 2 cents. I am trying to make a complete list.
The factors that affect ISO are:
The amount of light hitting the sensor. Less light more noise. You can at some extent control it manipulating your aperture and the shutter speed, or adding more light to the scene.
The quality of the sensor. Speaks for itself.
The area of the sensor. The bigger the lesser noise.
The working temperature of the sensor. The colder it is the less noise it has.
The area per sensel (pixel) here the Megapixels have a bit of impact, but this variable is compensated by the area of the sensor. If two sensors are of the same size, but a different megapixel count the lower megapixel will have less noise, but a similar result could be achieved resampling the bigger megapixel count to the lower one.
The Magnification of the signal. See point 7.
The computation made by the camera to compensate for it. This is where the ISO comes into play, the algorithm used to boost the ISO is calculated for the specific characteristics of the sensor. Here is one of the advantages of using more modern processors, besides more modern sensors. Another computation can be done is the averaging of a long exposure shot.
The amount of manipulation the user makes to the file, referring to how much boosting of the shadows is made using for example curves.
The post-processing done to a file, referring to a noise reduction algorithm.
Some other noise reduction technique like combining different shots. This is similar to the amount of light hitting the sensor (point 1) but done artificially.
Any resampling is done to the file. Resampling down a file with noise using an averaging algorithm like bicubic will smooth it. See point 5.