Running water, Cumbria, April 2008After RAW conversion
I have been pretty dissatisfied with long exposures using the digital camera. Anything past about 10-15s leads to horrid noise or detail loss with black frame removal. Plus a stack of filters seems to degrade the image more than film (or maybe that's just me).
Anyway, while processing the above image, I tried something different. I'd taken about 6 exposures of the scene with exposures varying from 1/3 to 1s. I was trying to get a smooth flow to the stream of water on the right. I only had a 2 stop ND filter, which didn't help. Looking through all the exposures I was considering pasting together parts from each to get the effect.
I then turned to Photomatix Pro to blend all 6. Not HDR, though. I just used the average function and some slight tone mapping. What a huge difference, seen here:
Running water, Cumbria, April 2008After the full Photomatix & post process treatment
Total exposure time for the 6 images was about 5s. The overall result though is of a much longer exposure. There is some great effect from the smaller flows amongst the rocks (100% crop below). Plus noise is gone and detail is great, something Photomatix is really good at.
Crop from final image, click for actual pixels viewSo there it is: take a whole lot of relatively short exposures and average the results. Output resembles a longer exposure than the sum of the parts.
I've not tried this yet with some really short exposures, but that would be fun. Say 10 or more at action freezing speeds (1/125 and faster).
Oh, and before you all chip in, I realise there were a bunch of film based techniques similar to this but I've not come across something similar in the digital world.
Posts
The statistician in me wants to point out that, for the static portions of the image, the noise should drop as a function of the square root of the number of samples taken, so that noise should decrease as you merge more and more frames.
ReplyDeleteAnd there's an application out there which uses the sub-pixel movement between frames to increase resolution as well.
Hmm...
Paul, I wasn't aware of the square root relationship but I stink at stats. Wouldn't it also depend on the near-pixel contrast definition being used for noise? Do I care? What I do like is that noise drops so I can also get deep into the shadows.
ReplyDeleteI know about the sub-pixel thingumie - super-resolution they call it. I think you mean Photo Acute. I get the feeling that Photomatix does something similar, just doesn't brag about it. What would such an app do with the moving bits?