factors that affect noise at higher ISO

[thepriceizrite]

I have a Canon PowerShot G10, which does not have great performance at higher ISOs. But I took two shots - the first at ISO 400 and the second at ISO 800 - and the second one had considerably less noise than the first, even though it was at twice the ISO. I've posted the images below:

1.



ISO 400
30.5mm
f/4.5
1/200 shutter speed
Macro zoom

2.



ISO 800
30.5mm
f/4.5
1/1,250 shutter speed

Both taken on a PowerShot G10 handheld (no tripod).


Notice that the first picture has more noise than the second, even though it is at a lower ISO. Is this because I was zoomed up close, and so that magnified all of the noise?

This got me thinking: what are the factors that create more noise in an image? I know higher ISOs always lead to more noise, but based on these two images, it looks like ISO is not the only factor that causes image noise (because here, the higher ISO shot is less noisy). If I have to use a high ISO when shooting, are there ideal conditions I should shoot in that would reduce the noise in the image?

Thanks!

 

[Overread]

Noise is also generated when you underexpose a photo. A photo underexposed and then brightened up in editing will show more noise than if you'd have exposed it correctly and used a higher ISO in the first place.

In addition if you're using a point and shoot camera and use "electronic" zoom then chances are its just magnifying the image and blowing it up (to simulate the effect) and as such yes the noise will become more apparent.


Interestingly webresize and print normally hides a lot of noise typically. Printing in itself can't typically render noise whilst web-resizing typically cuts a photo down a lot in size so the noise gets lost in the conversion (assuming you go for around 1000 pixels on the longest side).

 

[khakoo]

I think Overread has identified the problem in this case. The metadata reports a digital zoom factor of >1.7 for the noisier image and 1 for the cleaner shot. I take it "1" indicates no digital zoom engaged, or if not then less than the other picture anyway. Perhaps somebody will put me right if my reading of the factor is incorrect; it's a "feature" for which I've never had any use. You might want to check zoom options as your camera could be set to use digital zoom automatically upon reaching the maximum optical focal length.

Another cause of noise is a lengthy period in live view mode (when drawn from the primary sensor, as in most cameras) or a very long exposure, due to the image sensor warming up While energised. If live view must be used then it's best to power down the image sensor for a short time. It's hard to say how long this should be though. When I use a protracted period of live view I try to stop running the sensor for about a minute before shooting. Cooling time may not always be available of course, but any at all will help. Variables like ambient temperature and convection / conduction paths will also affect dissipation rate.

 

[Overread]

Khakoo you reminded me indeed of heat induced noise, mostly because I tend to only find it crops up when people speak of astrophotography (where shutter speeds into hours is not unheard of - although most practically use shorter exposures and combine the results)

Ambient temperatures do have an effect, though in general most use I think is marginal - its one of those things that has to go to the extremes to really become apparent; that said I don't think many have done studies (least personal ones) as to seeing the effects of shooting in different climate on noise alone

 

[TCampbell]

Did you shoot these as RAW or as JPEG?

If RAW then you'll see the level of noise that actually existed in the sensor.

But if JPEG... something special happens. The cameras have built-in noise reduction algorithms to deal with noise and when the camera converts to JPEG it applies these algorithms (it also applies other things... white balance, sharpening, "picture styles", etc.). But since the noise level correlates to the ISO setting, the camera changes the aggressiveness of noise reduction based on the ISO being used.

But this gets interesting because it doesn't actually change the noise for each individual ISO... it tends to go in groups. ISO 100, 200, and 400 might get no noise reduction. ISO 800 & 1600 might get moderate noise reduction. ISO 3200 & 6400 might get aggressive noise reduction, etc.

It's not usual to find that ISO 800 images have less noise than ISO 400 images *if* you shoot JPEG. If you shoot RAW you'll notice the noise correlates more closely to your ISO (and you'd apply noise-reduction using the computer.)

 

[thepriceizrite]

Thanks for the input everybody. It seems that the main issue here was my use of the digital zoom. My camera does have a digital zoom that sets in automatically when the lens has reached its maximum length.

I think Overread has identified the problem in this case. The metadata reports a digital zoom factor of >1.7 for the noisier image and 1 for the cleaner shot. I take it "1" indicates no digital zoom engaged, or if not then less than the other picture anyway. Perhaps somebody will put me right if my reading of the factor is incorrect; it's a "feature" for which I've never had any use. You might want to check zoom options as your camera could be set to use digital zoom automatically upon reaching the maximum optical focal length.

Could you explain to me what metadata report you are referring to? I am not familiar with this feature. To figure out what settings the photo was taken on, I just looked at the "info" bar in iPhoto, but I don't know any way other than that to find out the settings a picture was taken on.

Did you shoot these as RAW or as JPEG?

I shot them both as JPEG, which I always do. I've never seen an advantage to shooting RAW. Thanks for sharing that info about automatic noise reduction.

 

[khakoo]

Could you explain to me what metadata report you are referring to? I am not familiar with this feature.

There's a lot of very useful information stored with the image file a camera creates. Typically the camera model and lens, exposure mode and settings, focal length selected, focus distance, etc. Unfortunately there isn't yet an industry standard format for metadata. You'll see people refer to it as EXIF data, I've even done this myself for consistency in threads, but in fact EXIF is just one prominent attempt at a standard, and it too is inadequate. Manufacturers have implemented various methods of tagging files with information about the device and settings, and while there is a great deal of overlap, there's still too much incompatibility. The tag fields and values may or may not be visible depending on the software with which you view them, and some of the data can be lost, or stored incorrectly, if a file is re-saved from software that works differently from the original. The old rule applies as always: Keep your original file pristine and only work on a copy. There are also privacy issues with the detail and persistence of metadata so I strip all of it from the versions I display unless needed for a known purpose. In your case I found what I was looking for by the most rudimentary method of simply opening the file's Properties in Windows and going to the Details tab.

As for your image quality problems, I recommend you disable digital zoom entirely. If it's ever truly necessary to upscale an image it would still be best to do so on a computer rather than in-camera. Really though, I wouldn't do it at all, and it's quite disingenuous of manufacturers to pretend digital zoom is a legitimate substitute for genuine optical performance. If your zoom isn't long enough then shop for a camera / lens combo with sufficient focal length. If you want a large print then just output at a low enough dpi resolution rather than interpolate a higher pixel count.

I've never seen an advantage to shooting RAW.

There are tremendous advantages to RAW, especially over JPEG. Even if you don't want to get into it yet, storing your images in RAW form provides the most potential for your images, as it captures the entire sensor data of which your camera is capable, and offers the most scope for processing the way you want instead of accepting irrevocable and destructive decisions by the camera. Even the highest quality JPEGs permanently discard valuable image data, particularly tonal range. I think the G10 includes the option of making a JPEG at the same time as the RAW file when you shoot. Alternatively you can have JPEGs created from RAW on your computer. Either way you can have both file types if you wish, and storage capacity is so cheap now there's no need to be frugal.

 

[wfooshee]

Digital zoom is like cropping a larger picture. It does not take the picture and simply blow it up, resulting in the same resolution of a smaller section of your view. It reduces the amount of sensor being used, which causes two immediate effects. You amplify whatever's wrong with the section of the picture you're using, and you reduce the total resolution. Had you shot the first picture without digital zoom, then cropped it to that same size becaue the subject wasn't large enough, you would have had the same effect.

Another noise amplifier is boosting the level or exposure in an underexposed image. The noise, usually almost buried in the image data, gets boosted, too, and becomes much more visible.

Don't use digital zoom. If you HAVE to simply because you can't get close enough and the rest of the frame is garbage, just remember the increase in noise and the reduction in resolution.