How does the D3 achieve such high ISO?

Started Feb 17, 2008 | Discussions thread
ejmartin Veteran Member • Posts: 6,274
Re: 1D3 full well?

ohyva wrote:

As NR is typically decision controlled median type of local filtering
I very much doubt you can see that clearly in any fourier based
analysis. But I have to admit it's quite long time since my last math

OK, I took a 14-bit NEF file of a brightly exposed uniform wall at ISO 200

then split off one of the two green subarrays in IRIS and did the Fourier transform of a 512x512 pixel sample, both before and after applying a mild adaptive median noise filter (the command af3 .2 in IRIS). Here is the Fourier transform without any postprocessing noise reduction

A levels adjustment has been applied to focus on the high frequency noise which appears away from the center of the pattern, toward the edges. Note the uniform speckling away from the center, indicating that the noise spectrum at high frequency is flat all the way out to the pixel level. And now the Fourier transform of the same patch, with the same levels adjustment, after the mild adaptive median filtering

The darkening toward the edges is the sign of the dampening of the noise spectrum at high spatial frequency, due to the averaging over neighboring pixels (smearing due to noise reduction, in plain terms).

The fact that the noise spectrum of the raw data is flat all the way out to the highest spatial frequencies means that no noise reduction has been applied to the D3 raw data that involves any averaging over neighboring pixels. Noise reduction on shorter exposure times such as this one, does not happen on the D3.

-- hide signature --
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow