My Geeky Analysis of Noise and Night Photography

OK, If you are the creative type looking for some creative tips... skip this blog entry. If you like the technical aspects of photography, then you might find this interesting.
Warning.. this is pretty nerdy stuff.

After my first outing of night photography I was pretty happy with the results but concerned about the noise levels I was seeing in my captures. So I wanted to get a clear understanding of how noise impacts night photography. The specific questions I wanted to answer were...

• Does noise increase with exposure time?
• Is there a point where it increases dramatically?
• Does the Long Exposure Noise Reduction help a lot here?
• How bad is the noise really?

At first I was going to do a subjective evaluation of some test shots, and then as I started hunting around for a suitable test chart to shoot, I came across an ISO Standard test chart for evaluating noise (for the geeks it's called ISO15739).




I thought this looked promising, and then I came across a free plug-in for Photoshop that automatically performed the noise measurements from the standard test chart.

Click on any of the images to see the full-size image

Well that was pretty cool... all of a sudden my qualitative measurements could now become quantitative. Now if there are any ISO standards people out there who are going to have a go at me for my lack of controls blah blah blah.. for the record... I don't give a rats! It's good enough for me so don't waste your breath.

Doing a first pass to get used to it - The impact of ISO on Noise.
I thought I would start with something simple to get the hang of the process, so I shot a series of test shots of the chart at different ISO ratings. From 100 - 3200 with my D200. Here's the results of that test.

Nothing overly surprising to me... Just so you understand the chart, across the bottom (x-axis), it shows you the different luminance levels (there were 12 patches on the chart), so you can see the shadows on the left and the highlights on the right. The Y axis is the average noise that was read by the tool (I averaged the horizontal noise readings and the vertical noise readings to create one AVERAGE noise reading for each patch). It was interesting to see how the noise really goes out of control over ISO 400, and how @ ISO100 and ISO200 the noise is for all intensive purposes the same.

Test2 - Impact of Exposure time on Noise
Next test was to understand two things... the impact of exposure time on noise and also whether the Long Exposure Noise Reduction does much (except piss me off when I'm taking the photos and waiting for it to process). I took 2 exposures at 15sec, 30sec, 1min, 2min, 4min & 8min. One exposure with NR on, and one exposure with NR off. Now this wasn't a totally controlled environment (the sun was setting and light was changing, but I tried to get the histograms roughtly the same for each set. Here's the results.


I didn't graph all the results because the graph would be too confusing. The pink & yellow lines are the 15sec exposures, and the cyan and purple are the 8 min exposures. OK... so you are probably thinking "What gives" How come the 15sec exposures have more noise than the 8 min ones.... Basically that is because I exposed the 8min exposures with +1.5ev and the 15sec exposures were + 1ev.... So the 8 min exposures were "exposed to the right" of the histogram more and required less exposure boost in the RAW conversion to get the patch densities to the correct level. So two key takeouts here from this graph.
1) Exposure time doesn't appear to impact noise levels (at least up to 8min duration)
2) Long Exposure noise reduction does bugger all (it did stop a couple of red pixels, but thats it).


Test 3 - Impact of Under and Over Exposure on Noise
So I did one final test. I wanted to understand the impact of exposing to the right on noise. I shot 4 exposures at the same shutter speed - The first @ -1ev, then 0ev, then +1 ev, and finally +1.5ev. Note: Even at +1.5ev the histogram was not clipping, so this was definitely "exposing to the right" without clipping. Here's the results.


So my key learnings from two days of Geeky stuff.

1. Super long exposures don't create noise in my D200
2. Long Exposure Noise Reduction is not worth it (at least for exposures up to 8 min)
3. Under exposing (even slightly will give you the biggest noise hit)

Some of you experienced guys are probably reading that thinking "No shit sherlock"! Yeah I know I know... but at least I fully understand the dynamics of noise now on my camera.

Brent