Iso 25600 ?

[ajohnw]

I notice my EM-5 would go up to ISO 25600 a couple of days ago so took a dimish light shot into a corner of the living room.

Results are better than expected but no where near the usual ability to lift up dark areas and noise is apparent if viewed full sized. Focus was on the spikey ball and metering there mostly too. Noise has made the image fuzzy in places. Colours lost too in places. 100% crop.



John
-

[dje]

John was this an in-camera jpeg or raw ? If jpeg, does camera use noise reduction ? If raw, did you use noise reduction ?

Dave

[jcuknz]

I rate that as excellent for a 25600 shot completely acceptable for shots taken when one needs that speed. MFT is good

[ajohnw]

JPG and straight from the camera other than brightening up the little square. No noise reduction set. Nothing at all other than keep warm colours which is an Olympus thing they are famed for. Going on the blown out front to the white box there seems to be a large reduction in dynamic range. Low energy bulbs and the colours were slightly off so rebalanced on the white "fragile" tape.

I haven't looked at the raw yet but will later today.

I think I will get the tripod out tonight and shoot a series reducing ISO on the basis best to find out before actually needing to use it..

John
-

[dje]

JPG and straight from the camera other than brightening up the little square. No noise reduction set. Nothing at all other than keep warm colours which is an Olympus thing they are famed for. Going on the blown out front to the white box there seems to be a large reduction in dynamic range. Low energy bulbs and the colours were slightly off so rebalanced on the white "fragile" tape.

I haven't looked at the raw yet but will later today.

I think I will get the tripod out tonight and shoot a series reducing ISO on the basis best to find out before actually needing to use it..

John
-

Thanks John, interested to see how raw compares ie whether in camera jpeg is applying noise reduction without you knowing !

Dave

[DanK]

I don't know anything about that camera, but the ones I am familiar with apply NR to in-camera jpegs. If the manual does not say, a similar raw file with no NR would answer the question. Since you will be shooting it raw, pls let us know what you find out. Also, a link to a larger image would be helpful, because noise is harder to see a low resolutions. (I'm viewing on an iPad, so my apologies if you posted it large and that just doesn't show up on this device)

[shreds]

Whilst low light/high ISOs have revolutionised in the past few years, I have to say that I would be reasonably pleased with that outcome. It is never going to be perfect, but it nevertheless achieves a result.

In much of the professional work that I do, a result rather than an artistically and technically perfect result is what is necessary for the client. So this compares well.

Be interested on the RAW results when you have done a bit more pp.

Softness and blown highlights are usually unavoidable in such a situation, however.

[ajohnw]

The images I have posted are all 100% resolution. On this forum clicking on an image shows it in a lightbox and clicking it again on the small square takes it up to original size. Haven't tried it on an iPad though.

This is from raw. Straight., no curves use or anything else. The noise as such is mainly shown as black clipping during conversion. There is also slight white clipping in the highlights on the knight.



I'm prepared to accept that this isn't the usual in camera noise reduction. More some specific processing related to ISO settings. The raw file has kept the colours but they are clipped all over the place so where this happens correct colour can't be seen only splodges of colour. The white clipping is different in the jpg so shows some highlight compression.

It's interesting to note that in some ways the raw conversion is "better". This is down to the use of a linear conversion curve rather than the compress both ends used on many cameras these days. Nikon for ages and Canon only recently.

So far attempts to remove the noise loose contrast even noticeable when reduced to small size 1000 odd by etc. Wavelet reduction seems to be the most effective. Different debayering techniques might help a bit too.

John
-

[dje]

Looks pretty good for ISO25600 !

Dave

[ajohnw]

I tried a bit of noise removal. Just had to try Rawtherapee. It uses an unusual method that seems to work well but I doubt if I hit optimal settings as have never used it before. These are 50% reductions of the others. Full sized slight colour noise would be visible.

Original just reduced 50%



Rawtherapee noise reduced. I'm impressed that it has got rid of that much without much blurring/



That image sharpened at about 50% of the usual max setting. Probably only see the difference enlarged.



The extra processing was done on the jpg from raw.

John
-

[DanK]

Thanks for pointing out how to view the full-sized images.

The raw file has kept the colours but they are clipped all over the place so where this happens correct colour can't be seen only splodges of colour.

You lost me there. I don't follow that this has to do with clipping. The image shows typical noise--a random graininess. It's quite apparent on areas lacking detail, like the side of the box. Still, it is not bad for such a high ISO.

I have never used Raw Therapee, but if you process in Lightroom, the noise reduction since version 4.0 has been quite good.

[ajohnw]

I can't post a picture as the clipped areas flash but I can assure you that they are clipped. On 2nd thoughts I can as flashing is optional in Rawtherapee. The white dot's have a channel at less than 2 bits. Some are in odd places. I used ufraw to develop the raw. Goin on other work I have done with this it does show true clipping. There was white clipping as well. Overall I feel it's usable for images that don't have a large dynamic range.



The only problem with Rawtherapee really is that some work has to be exported to the GIMP. Dodging, burning and cloning with it are easy. Some techniques need layers and masks - more than Colin uses on PS in some cases to do the same thing. As I learn more about it I find layers can be used to make it none destructive as well. I never alter the original image anyway and what ever is done can be saved in GIMP style files. I assume there is also a macro recording type plugin for it as well so that many images can be batch processed.

John
-

[rpcrowe]

It's a very wise practice to investigate the maximum ISO that you will accept from you camera BEFORE you need to shoot something important that needs a high ISO!

[ajohnw]

It's a very wise practice to investigate the maximum ISO that you will accept from you camera BEFORE you need to shoot something important that needs a high ISO!

Exactly. That's why I tried it. Must admit it's better than expected. I need to do a number. On this shot it will change exposure mode when the iso is dropped to some level.

John
-

[Downrigger]

Does anyone know whether high ISO has a different (nature, amount) of noise when shooting dim light as opposed to bright light with fast shutter speed to stop action? Maybe a dumb question - the obvious answer being "it doesn't" - but obvous and correct are more often different than I'd have thought.

[MrB]

I'm not sure what I am looking at here - what is the little rectangle just above the white box?

Philip

[ajohnw]

The small box is an area that I have selected and brightened to see what is buried in the shadows. The camera has taken some of it to grey in the camera jpg shot. There are all sorts of if's and buts in camera firmware jpg processing.

The usual main effect of higher ISO settings is less dynamic range and more noise the darker the shot is / longer the exposure. The only way to find out really is to try it. The lighting in these shots is lower than what might be used in say a night club etc on stage or an indoor event of some sort.

John
-

[MrB]

Then I don't understand the apparent lack of noise in these images (viewed full-size in the light box), when compared with the Imaging Resource test image at ISO 25600 (also viewed full-size), shot in a studio with powerful studio lighting (50mm f/8 1/2500s).

http://www.imaging-resource.com/PROD...LI25600NR0.HTM

Philip

[DanK]

I think part of the difference may simply be the reduction in size. John wrote that his is a 50% reduction, but I think it is more. The image to which you linked is approximately 95 cm on the vertical dimension on my screen. John's original as posted is 32 cm. Compressing to 1/3 of the original size should reduce the impression of noise quite a bit. However, I don't use Raw Therapee, so I have no idea what other processing it is doing in its default rendering.

Then I don't understand the apparent lack of noise in these images (viewed full-size in the light box), when compared with the Imaging Resource test image at ISO 25600 (also viewed full-size), shot in a studio with powerful studio lighting (50mm f/8 1/2500s).

http://www.imaging-resource.com/PROD...LI25600NR0.HTM

Philip

[ajohnw]

As I said the only way to find out is to try it. Somehow I don't think you have looked at the straight from raw image I posted at it's maximum size.

This is the uncropped one Edit LOL I thought it was but it's actually the cropped one. I don't have the original full sized one Yes - that's what i have done cheated. If any think that tough luck.

http://www.23hq.com/ajohnw/photo/14725541/original

On the other hand very roughly denoised, reduced to 50% and sharpened.



John
-