The Great Noise reduction challenge!

[Overread]

OK, did the experiment and Tsarex is exactly right. Even though the light is sufficient and the same for each shot, the noise jacks up as you crank up the ISO. At 6400 it was not that noticeable but at 12000 and HI it was very noisy.

Exactly and this is where the "always keep the ISO low" theory is born. Or rather the "keep it as low as the situation allows" concept for ISO. Since raising it always triggers an increase in noise. The key is that raising it always delivers a cleaner shot than using a lower ISO and underexposing.

 

[Ysarex]

OK, did the experiment and Tsarex is exactly right. Even though the light is sufficient and the same for each shot, the noise jacks up as you crank up the ISO. At 6400 it was not that noticeable but at 12000 and HI it was very noisy.

Exactly and this is where the "always keep the ISO low" theory is born. Or rather the "keep it as low as the situation allows" concept for ISO. Since raising it always triggers an increase in noise. The key is that raising it always delivers a cleaner shot than using a lower ISO and underexposing.

I'd say that has been the long standing correct assessment, but we are seeing some recent changes in the tech that may require us to add a qualification. A new term is showing up as some newer sensors are being referred to as "ISOless." Give it a Google.

It only has minor implications for RAW ONLY SHOOTERS and is otherwise interesting from the standpoint of understanding how the bleep thing works.

Joe

 

[Overread]

Certainly with the continued improvement of ISO noise and detail levels we might well enter a time when too little light becomes increasingly less and less a problem for getting quality photos. Especially for web and print display.

The underlaying theory will still remain the same, we might just enter a time when the difference at output between ISO 100 and ISO 1600 is impossible to tell apart (or even higher ISOs!)

 

[TCampbell]

Here's my version processed with Aperture and Noiseware Professional:

What I like about Noiseware Pro that I don't get out of many of the others I've tried is the ability to "tune" the type of noise and the aggressiveness of dealing with the noise and I can do this by tonal frequency in the image. Typically I'll notice stronger noise in the shadows... weaker noise in the highlights. But this image needed a lot of de-noising everywhere so I didn't fuss over the tuning.

Also... since the image was going to be down-sized to 1024 pixels on the longest edge, and resampling to a smaller size has the side-effect of de-noising anyway... I reduced the aggressiveness of Noiseware to avoid over-softening the image. (Over-sharpened images tend to look like "plastic" and they lose the organic look of a real subject.)

Lastly, I did want to recover some sharpening. But sharpening and de-noising are basically enemies. So I didn't want to apply global sharpening to the image (as this basically fights to undo the de-noising by re-introducing more noise). Instead I painted in the sharpening in only the areas where I felt it would help. Namely the horse's head; the rider's head; the horse's tail; some saddle detail; hooves; and the dirt texture on the floor. The wall in the background and most of the horse's body didn't get any sharpening applied.

No other adjustments were applied (e.g. no color, exposure, etc. since this is supposed to be all about the sharpening.)

 

[dennybeall]

Very nicely done. A recovery technique for problems of this type is a very useful tool.

 

[bribrius]

We talk about taking photos with the ISO set at a high number, like 6400 or higher, but we also need to keep in mind about the amount of light available to capture the shot. We know the noise comes in if the LIGHT IS LOW and the ISO is needed to capture the information but what if the light is sufficient, do we still get noise?

not as much. just speaking from practical experience. I shoot higher iso daytime with limited light I will deal with much less noise than higher iso with even less light. I don't think it is necessarily LESS noise but often how much is visible. Underexposing and shadow recovery both bring out more. Shooting low light daytime higher isos it seems you would be less likely to run into a noise issue. least in my experience. if it were just iso that made the noise apparent then I couldn't compare two different photos with the same iso and have two diferent detectable amounts of noise. course shutter speed differences and exposure time with in camera noise reduction can often taint the samples.