Does low light performance improve if you lower resolution?

[RedTele]

If you set up the camera to take lower resolution photos, does it take better cleaner low light photos because you have lower pixel density?

You have more sensor area per pixel output, so does this mean you can use lower ISO than normal? Or same ISO with higher shutter speed?

I have a feeling it doesn't quite work like that...

 

[boyfalldown]

If you set up the camera to take lower resolution photos, does it take better cleaner low light photos because you have lower pixel density?

You have more sensor area per pixel output, so does this mean you can use lower ISO than normal? Or same ISO with higher shutter speed?

I have a feeling it doesn't quite work like that...

Shutter speed will stay the same for a given ISO and aperture (assuming nothing else changes) regardless of pixel density.

It does mean (in theory) that you can reach higher ISOs with a lower pixel density

 

[RedTele]

I thought ISO was sensor sensitivity. So if you have more sensor area per output pixel, you don't have to expose it as long to get the same amount of light, no?

 

[Suvv]

Changing the ISO just changes the amplification of the signal from the sensor, it does not change the sensor sensitivity itself.
When you shoot at low resolution the camera still uses all the pixels so the noise is the same in each pixel regardless of what resolution you choose.
At lower resolutions the software in the camera combines the data from adjacent pixels to create the new low resolution image. The combining algorithm can improve the signal / noise ratio thereby reducing noise in the final image.

 

[RedTele]

That makes sense. So unless the camera's in-built noise reduction is better than whatever noise reduction software you normally use, there's no benefit in shooting at lower resolution, apart from saving space on the card. Correct?

 

[Suvv]

That makes sense. So unless the camera's in-built noise reduction is better than whatever noise reduction software you normally use, there's no benefit in shooting at lower resolution, apart from saving space on the card. Correct?

The only other slight benefit is if you are on continuous shooting, the camera's buffer will take more shots at the lower setting. Whether it will make any practical difference depends on your camera & card, I can fill my buffer on raw but not on any of the jpeg settings.

 

[boyfalldown]

Changing the ISO just changes the amplification of the signal from the sensor, it does not change the sensor sensitivity itself.
When you shoot at low resolution the camera still uses all the pixels so the noise is the same in each pixel regardless of what resolution you choose.
At lower resolutions the software in the camera combines the data from adjacent pixels to create the new low resolution image. The combining algorithm can improve the signal / noise ratio thereby reducing noise in the final image.

Thats not quite right - by having larger pixels to start with (a lower pixel density) you get a higher signal:noise ratio to start with, so the noise at each pixel isn't the same regardless of pixel size/density

 

[Suvv]

Thats not quite right - by having larger pixels to start with (a lower pixel density) you get a higher signal:noise ratio to start with, so the noise at each pixel isn't the same regardless of pixel size/density

But we are not talking about larger pixels, we are talking about changing the resolution i.e. Large or Small.

 

[boyfalldown]

But we are not talking about larger pixels, we are talking about changing the resolution i.e. Fine JPEG versus Normal JPEG for example.

from the OP

does it take better cleaner low light photos because you have lower pixel density?

so I read that differently from you

but as you say it will make now difference what soever if you change compression algorithims in camera

 

[Suvv]

from the OP



so I read that differently from you

but as you say it will make now difference what soever if you change compression algorithims in camera

I edited my post that you quoted at the same time as you were typing your reply!
I was talking to someone about the difference in Size (resolution) & Quality (compression) and when I replied on here I mentioned Quality when I meant to mention Size

 

[RedTele]

I was talking about the settings in your camera that you can change i.e. Raw vs sRaw or Large vs medium or small JPEG. Just trying to figure out if I can get more out of my current camera.

 

[RedTele]

The only other slight benefit is if you are on continuous shooting, the camera's buffer will take more shots at the lower setting. Whether it will make any practical difference depends on your camera & card, I can fill my buffer on raw but not on any of the jpeg settings.

Yes I did wonder about that. However, processing (resizing) image is an additional task for the camera's processor to handle so wouldn't that make it slower, or is the bottleneck purely due to buffer capacity and card write speed?

 

[boyfalldown]

I was talking about the settings in your camera that you can change i.e. Raw vs sRaw or Large vs medium or small JPEG.

in that case as Suvv they have no effect on noise and only effect the way the camera processes images after they are captured

 

[Suvv]

Yes I did wonder about that. However, processing (resizing) image is an additional task for the camera's processor to handle so wouldn't that make it slower, or is the bottleneck purely due to buffer capacity and card write speed?

Yeah or No
Not sure to be honest, its probably easiest to try some and see as it is so dependent on the camera & card !

 

[Howi]

If you set up the camera to take lower resolution photos, does it take better cleaner low light photos because you have lower pixel density?

You have more sensor area per pixel output, so does this mean you can use lower ISO than normal? Or same ISO with higher shutter speed?

I have a feeling it doesn't quite work like that...

Your two statements don't quite add up, you can set your camera to use lower resolution, but that does not mean the pixels will have more area.
The size (area) of a pixel will be determined by pixel density on any given sensor.
for example a 1" square sensor with 10mp will have a worse level of noise than a 1" square sensor with 5mp (all other things being equal)

a lot will depend on the resizing algorithm as to whether you will see more or less noise due to the lowering of the Jpeg resolution - it can in theory work both ways.
You have to way up whether the (possible) gain in noise reduction outweighs the loss in other areas.
Noise reduction is an ever changing science, I have no desire to resurrect the RAW v JPEG argument, but, for me, shooting RAW only, means I can use noise reduction software as and when it is released, rather than relying on the camera firmware.
A simple example would be LR 2 and LR 3, the difference in noise reduction is staggering, it may no be perfect, but shows how software is progressing all the time (as is camera firmware to be fair), but you need to get a new camera to take advantage, you can't do it retrospectively.

 

[HoppyUK]

The total amount of light falling on the sensor would not change.

 

[RedTele]

If you set up the camera to take lower resolution photos, does it take better cleaner low light photos because you have lower pixel density?

You have more sensor area per pixel output, so does this mean you can use lower ISO than normal? Or same ISO with higher shutter speed?

Your two statements don't quite add up, you can set your camera to use lower resolution, but that does not mean the pixels will have more area.
The size (area) of a pixel will be determined by pixel density on any given sensor.
for example a 1" square sensor with 10mp will have a worse level of noise than a 1" square sensor with 5mp (all other things being equal)

Sorry, what I said in my second paragraph is what I meant. You have more sensor area used per output pixel. It's where you set the frame of reference. Different from conventional meaning of pixel density I shouldn't have said that.

 

[magicaxeman]

No, because the pixel size remains the same.

The reason lower pixel densities work better for low light is because lower density sensors use larger pixels on the sensor, larger pixels = more surface area to capture photons = more light captured.

All dropping your resolution does is lower the number of pixels that are read from, it doesn't double up and read two pixels information and count it as one pixel (which would give an effective increase in the surface area that can capture photons.)

With sensors you need to think of them as a grid of small square sensors, each of which is a few microns across and not "sensor" as a whole

 

[RedTele]

No, because the pixel size remains the same.

The reason lower pixel densities work better for low light is because lower density sensors use larger pixels on the sensor, larger pixels = more surface area to capture photons = more light captured.

All dropping your resolution does is lower the number of pixels that are read from, it doesn't double up and read two pixels information and count it as one pixel (which would give an effective increase in the surface area that can capture photons.)

With sensors you need to think of them as a grid of small square sensors, each of which is a few microns across and not "sensor" as a whole

So what you're saying is post #4 by Suvv is incorrect. What actually happens is that only ~% of the total pixels are activated. Say, every other pixel site is turned off, then the resulting image is 1/4 the max size, but because the size of each pixel site is the same, it doesn't affect the noise quality, despite the pixel sites being further apart from each other. Your explanation makes sense as it's far simpler for camera to produce a resulting smaller image.

 

[boyfalldown]

So what you're saying is post #4 by Suvv is incorrect. What actually happens is that only ~% of the total pixels are activated. Say, every other pixel site is turned off, then the resulting image is 1/4 the max size, but because the size of each pixel site is the same, it doesn't affect the noise quality, despite the pixel sites being further apart from each other. Your explanation makes sense as it's far simpler for camera to produce a resulting smaller image.

No that's not right. Size options in this case are determined by the cameras processing and not the sensor, so if I choose to take low quality small jpegs then I'll still take a 12mp photo but the camera will compress the resulting photograph far more aggressively then if I choose to take the same photo using a large high quality jpeg, but the initial photos in both cases will be taken using the same 12mp

(if that makes sense)

 

[RedTele]

lol ok then! Yes I understand what you both are saying. Just don't know which is the actual process happening!

 

[boyfalldown]

As far as I'm aware the only way to switch parts of a sensor off selectively is by using some of the crop options on FX Nikons. Switching individual pixels off in the way suggested above would represent an immense technical challenge

 

[Howi]

You wouldn't need to switch pixels off, just not read them.
If you reduce the resolution say from 8mp to 2mp (by selecting small jpeg) the processing software would only need to read every fourth pixel.
I would imagine, as this is the simplest aproach, that is what happens.
A more complex way would be to select every fourth pixel and compare it to it's immediate neighbours and take the average, but I still think it will make no difference to the noise.

 

[boyfalldown]

You wouldn't need to switch pixels off, just not read them.
If you reduce the resolution say from 8mp to 2mp (by selecting small jpeg) the processing software would only need to read every fourth pixel.
I would imagine, as this is the simplest aproach, that is what happens.

Surely the simple way would be to take the raw image and apply your jpg compression algorithms to it with compression and sizing options set as the user has chosen in camera. (like the way its done)

 

[Howi]

Surely the simple way would be to take the raw image and apply your jpg compression algorithms to it with compression and sizing options set as the user has chosen in camera. (like the way its done)

it's the same difference really, whether you do the processing with the RAW file or direct from sensor, to reduce resolution you throw pexels away ( or ignore them). But, as you say, the way it's done (I suspect in most if not all cameras) is to produce the RAW file, throw away pixels not needed, then do rest of conversion to Jpeg.

 

[HoppyUK]

Sony has a range of compacts using their EXR technology, like this new one http://www.dpreview.com/news/1101/11010512fujifilmf550f500exr.asp

It does some clever things like cutting the resolution in half and effectively giving adjacent pixels a different ISO level, which extends the dynamic range quite significantly. And it works well.

They make some other claims which sound more like marketing speak than anything else. As far as noise is concerned, the limit is set by the total number of photons the sensor can capture, the amount of amplification that is needed to deliver a well exposed image, and the efficiency of the processing engine and software.

 

[Suvv]

it's the same difference really, whether you do the processing with the RAW file or direct from sensor, to reduce resolution you throw pexels away ( or ignore them). But, as you say, the way it's done (I suspect in most if not all cameras) is to produce the RAW file, throw away pixels not needed, then do rest of conversion to Jpeg.

You do not throw pixels away or ignore them, you use all the original pixels to create fewer new pixels.
As noise is random it is possible for the noise to be reduced when this resampling takes place.

 

[Hodders]

You do not throw pixels away or ignore them, you use all the original pixels to create fewer new pixels.
As noise is random it is possible for the noise to be reduced when this resampling takes place.

This is what would surely happen. ISO performance is so competitive that not using noise averaging would just be daft.

 

[Howi]

there is a process called 'pixel binning' that works at the hardware level (output from sensor) that WILL improve noise performance, but at the expense of resolution i.e reduced pixel count, usually by a factor of 4 but this can vary dependant on how many pixels are taken to generate a new pixel)
As most cameras apply the size (pixel) reduction during the processing of the JPEG image from the RAW file, the reduction of noise (if any!) will not be the same as the hardware pixel binning process.
I believe Canon(on some models) have an option for small RAW (sRAW) files and would imagine this is done using pixel binning process.
The noise reduction process cannot take place (in the same way) once the RAW file is produced.
The quality of noise reduction programs these days is such that pixel binning post RAW is probably a waste of time as you are throwing away a lot of information to get some 'apparent' lowering of noise.
Noise reduction post RAW, if applied too agressivley, will always lead to loss of detail anyway, as it is difficult to determine what is noise and what is detail.

So! to try and answer the original post, Yes! it can, if applied at the hardware level but with considerable loss of detail (resolution), personally, not convinced it works particularly well(if at all) post RAW.

A MF camera that has 65mp native resolution was reduced to 15mp after pixel binning, makers considered 15mp with ability to shoot higher ISO for same level of noise was still an acceptable resolution for commercial work.(nice for some eh!)

 

[magicaxeman]

You wouldn't need to switch pixels off, just not read them.
If you reduce the resolution say from 8mp to 2mp (by selecting small jpeg) the processing software would only need to read every fourth pixel.
I would imagine, as this is the simplest aproach, that is what happens.
A more complex way would be to select every fourth pixel and compare it to it's immediate neighbours and take the average, but I still think it will make no difference to the noise.

This is indeed the method, nothing is switched off, its just not read, that process is controlled by software and easy to write in.

What its not doing though is combining the output from 2 or more pixels to create one larger pixel, you actually need physically larger pixels.

The pixel size on the 550D is 4.7 microns
The pixel size on the 5Dmk2 is 6.4 microns
The pixel size on the 5D is 8.2 microns

An object 8.2 microns across will obviously have more light fall onto it than and object 4.7 microns across.

Now we get to the complicated bit.
Canon have worked around the problem with smaller pixels by increasing the coverage the micro lens above them gives.

On the 500D for instance the micro lenses have gaps between them that receive no light, on the 550D they implemented a gap-less micro lens system that means each pixel receives more light than it did in the 500D.

Low light performance in the G range of compacts was not that good with the G10, for the G11 they dropped the pixel count to 10mp so you had larger pixels, this improved the low light performance, but I think it may have been further improved by gap-less technology on the G12 though I'm not sure as I haven't checked yet.

Then we have improvements in firmware and processing, these can help in pushing up the effective ISO of the camera.
I can still remember when a good camera had just ISO 50-400, its been now pushed to awesome levels, even my lowly 500D will give fairly decent shots @ ISO 3200, even then it can go on to 12800 if needed.
Later models can really push the ISO and good shots at ISO 6400 and higher are possible, all this is down to the better sensors and processing available in these models.
Another improvement here is how the newer sensor/firmware combination handles noise in images, most now automatically take and subtract a dark frame, its my betting it wont be long before they are taking flats to help combat any distortions in images, in fact its already probably being done in camera.

Pixel binning is what I was talking about earlier, its the effective combining of the output from a group of pixels and treating that output as the output from 1 pixel.

With 2x2 binning an area 2 pixels by 2 pixels has its charge combined and is treated as 1 pixel, so in effect increasing the surface area of a pixel by 4, whilst this lowers the resolution of the sensor it has the benefit of increasing the amount of light detected, improving the signal to noise ratio and lowering noise.

Camera's really have gone way beyond the simple amount of light falling on an emulsion and the choice of good fast lenses, they really are in the realm of rocket science as its astronomy (digital astronomy) which has led the way in how to maximise image resolution in low light.
I bet a lot here wont realise that a lowly 8mp astronomy camera can out shoot any DLSR when both are connected to the same telescope.

So yes you will get better low light performance from a lower pixel density as it will have larger pixels, but... the 5DMK2 will still compete or out shoot the larger pixeled 5D due to its gap-less micro lens technology and better processing.