Digital camera: question about noise

[newshound]

The bigger the sensor, the bigger the individual pixels (at a given megapix). The bigger the pixels, the lower the noise at high ISO.

My Canons has a setting which alters the x-y resolution of the image: L, M1, M2, M3, and S on the G10, for example. And another setting which also affects file size: Normal, Fine, and Superfine, which I guess might relate to the JPEG compression.

Now obviously, a low megapix file has lower ultimate resolution than a high one, but will save faster, and you can get more on a memory card.

My question is, if you select the lower resolution settings, does the processing do some combination of the "raw" data and will this reduce the noise on low light / high ISO images? From some not very scientific trials, I'm not convinced of any benefit. So is there any point not using L and Superfine all the time (if you are not trying to shoot fast sequences, and have plenty of memory, and are happy to post-process for web use etc).

No doubt this is discussed somewhere on the web, and I have tried looking but there is a *huge* amount of "noise" out there. Can anyone suggest any good links or "live" newsgroups? alt.photography doesn't seem bad.

TIA

Steve

[polygonum]

In my experience there is rarely any reason other than storage space for using anything but the most pixels and the least compression.

The noise issue relates to the physical sensor pixels so you would not expect any improvement from most settings other than the ISO. The difference occurs when you compare two cameras with markedly different physical sensor sizes.

In these days of multi-gigabyte storage media even storage is not much of an issue.

[meow2222]

Yes & yes, but only very basic. Postprocessing in gimp's always better IME.

no

[The Natural Philosopher]

Yea, even unto using 'raw mode', if the camera has it. Someone told me, and I believe it to be triue, that the raw images is a lot better than 8 bits deep and you canm sometimes reciocvver acceptable pictures if you process the raw file where the JPEG is 'clipped'

I think that is not what he is talking about. high gain on the CCD tens to introduce random noise into the image. TO an extent reducing te reoslutin will avrege out te 'final' pixle frtom teh adcabnet ones as well.

well I dunno. I put 250GB on the home server and the wife had filled it with random TV recordings in 3 months.

[rbel]

As a starter you may want to have a look at the entries in the Digital Imaging section of the DPreview glossary at

specifically those on Compression, Noise, Resolution and Sensitivity.

[polygonum]

Indeed re noise - but at least most of what a camera can do is also achievable with post-processing on a computer.

Too true 250GB can easily be filled! But my first digital camera had something silly like 16 MB cards... (or was it even less?) - and at least 4 GB lasts a bit longer even with huge image file sizes, and is cheaper!

[The Natural Philosopher]

Not if you already lost the data, The point is that te camera takes way? 3x12bit? 3x16 bit and slaps it into 3x 8 bits. Sometimes it simply loses some bits.

yes phots are not as bad as videos. 4Mbyte as against 4Gbyte..

[John Rumm]

The short answer is that its unlikely to be any better than what you could do yourself in post processing. Generally, newer cameras (especially DSLRs) introduce less noise at high ISO settings than older ones, but obviously the noise still goes up with sensitivity.

Downsampling an image will tend to smooth out the noise a little, but you can normally do better in post processing by using median filtering or some of the photoshop noise reduction filters.

Another trick that can be used if you take multiple exposures of the low light level scene, is to stack them up in layers, and blend them together. Since the noise is essentially random, it does not "add" in the same way the actual wanted picture information does, so you get noise reduction due to oversampling.

[djc]

processing do some combination of the "raw" data and will this reduce the noise on low light / high ISO images? From some not very scientific trials, I'm not convinced of any benefit. So is there any point not using L and Superfine all the time (if you are not trying to shoot fast sequences, and have plenty of memory, and are happy to post-process for web use etc)

Converting from raw to jpeg will reduce the bit depth, so information is lost, jpeg compression loses information because one value is used for several adjacent pixels, scaling down the image size will average out the value of adjacent pixels. If there is not much noise, it will be smoothed away by all this averagin out, if there is a lot of noise then it will have more influence and tend to blur away the detail. Not all raw to jpeg conversions are the same, nor all scaling algorithms, so cameras will vary, but however it's done information will be lost.

[Tim Streater]

Not nesser-celery. RAW is stored as at least 12 bit depth, sometimes 14.

[Martin Brown]

The only down side is you run out of media faster.

rec.photo.digital isn't bad.

Raw mode is typically about 12-14bits deep and so an N Mpixel image occupies about 1.6N MB. A highest quality JPEG is usually around N/2 MB worst case. Some cameras will save both for you. Saving speed of raw can be an issue if you are doing multiple frame action shots. It is better if getting optimum signal to noise is important to you.

Main advantage of raw is when you know you are facing one time subject material with insane contrast and highlights and shadow detail both matter. Wedding with bride in white and groom in black velvet being the most common serious dynamic range challenge. In camera JPEG default will almost certainly ruin the image with blown highlights.

The only time it is worth using less than the full resolution of the equipment you are carrying is when you are certain you will never need a higher resolution image (eg photo for web/eBay selling) or you would otherwise run out of storage media and have no blanks with you. Given the relatively low cost of sD cards this is not really an issue today.

It is possible that some cameras do bin the data down when certain magic ratios are selected (notably saving at 1/4, 1/9, 1/16 sensor size). Astronomical CCD cameras call this binning and do it to reduce the readout noise. Some of the newest cameras also take dark frames. (ie linear factors of 1/2, 1/3, 1/4)

Some cameras also take their own dark frames on time exposures and do a certain amount of in camera correction for dead/hot pixels etc.

But you can always trade resolution for signal to noise later by low pass filtering in the post processing chain. You cannot however get back the pixel data that was thrown away and never stored.

Stored content will always expand to occupy all the space available.

[polygonum]

Not for me!

Mushrooms. With stark white parts, some (almost black) parts, and lit by a mixture of direct sun, filtered sun and, sometimes, flash.

Amazing what can be dragged out of the shadows. But overexposure, whilst very easy to achieve, is unmanageable.

[Mark]

IMHO you will get the best images if you take them in raw mode and then compress them yourself on a computer. Of course this takes time and most cameras do a decent job of compressing.

Of course getting the correct exposure is the best way :-)

Pictures from a high end camera will generally have less noise in low light conditions, owing to a better sensor.

[The Natural Philosopher]

Yep. It wasn't a wedding, but you are close.

I am going to try raw mode and see.

Which is in the end the salient point.

Jpeg is always a more or less lossful compressions. Something is always lost. Raw mode is a humongous bit map but one in which you can select what you want to discard, and how you want to compress it, as part of the post processing.

Only practical downside is sped of picture storage as you said.

That goes for shoes as well.

[The Natural Philosopher]

sorry, I thought it was obvious that I meant 'when not on raw mode'

[The Natural Philosopher]

Have you tried raw mode and deliberate underexposure?

[Mentalguy2k8]

Agreed, I only use the very minimal amount of "camera post-processing" and the max amount of file size whether it's a digital SLR or a smartphone. Always select the biggest number of pixels and the least amount of compression that you can work with.

It's far easier to enhance a huge raw & uncompressed photo in photoshop etc., than to try and rectify all kinds of buggerations that the camera has "automatically" processed an overly-compressed small picture with. If the camera compensates for low-pixel photos, disk space or various other scenarios *before* you import it to PC, you invariably lose something and you can never get it back.

[polygonum]

Yes - and that is the way to go. I have actually been amazed how well some shots have turned out given the difficult lighting conditions. I prefer the most "natural" lighting possible so only use flash if essential.

[Mark]

And often using raw mode too. To avoid such issues you may need to use HDR techniques.

[The Natural Philosopher]

The problem I found as a user of manual and semi-automatic (AP) cameras for years, is that digital autos replaced one thing (a focus ring, and a split image rangefinder with in my case (Nikon D200) IIRC three buttons and switches with about 8 possible settings on each that gave several thousand options for focussing mode.

Likewise exposure is no longer 'center weighted' allowing you to swing the camera and see what the exposure would be for various parts of the frame, and them manually slect the best compromise. You have about 8 sensors if memory serves..various combinations of which can be combined to give the overall exposure. Now in 'idiot mode;' all this makes for, on average, better pictures taken by idiots (of idiots) but I am not altogether sure it makes the job of setting the thing up easier for the semi-skilled, which is about where I am at.

After a couple of months of out of focus pictures I finally dragged out the manual to find that some how the thing has been set to focus on 'top right' . I've set to focus on 'dead center' but of course if two people are standing apart that gets you a neatly focussed background....:-)

It's a bigger learning curve than an old manual camera. I remember guessing distances, depth of field ,and exposure, all..and frankly, getting it as right as the auto 9 times out of ten.

Raw mode is the next thing to explore. Trouble is I dont think SWMBO has the software to deal with it.

So it will be like leaving the toilet seat down.