photon statistics and LibRaw

Submitted by rawhide on

I am trying to generate a photon transfer curve from DNG data from an iPhone. Since I am interested mostly in pixel statistics, what are the most "minimal" processing parameters I should be using?

Since I'd like to obtain data for the 3 (or 4) CFA channels separately, I'm currently using:

postprocess(gamma=(1,1), no_auto_bright=True, no_auto_scale=True, output_bps=16)

and then running statistics on the individual array slices.

I'm finding the no_auto_scale parameter in particular strongly affects the measured standard deviation. Are there any other parameters you would advise I include, or exclude?

Thanks in advance!

Forums: 

You do not need

Submitted by lexa on

You do not need postprocessing step(s) if you want to analyze unaltered raw data

-- Alex Tutubalin @LibRaw LLC

You can use 4channels sample

Submitted by lexa on

You can use 4channels sample to split source RAW into four separate TIFF files.

-- Alex Tutubalin @LibRaw LLC

To add to what Alex just said

Submitted by ib on

To add to what Alex just said, what demosaic, when what you want is sensor PTCs?

--
Iliah Borg

Well, I was originally trying

Submitted by rawhide on

Well, I was originally trying to perform my analysis on difference images (subtract one image with identical parameters from another in order to remove flat field effects) but I'm getting ridiculously large values for the std dev (in the 1000s) which doesn't seem right. Am I being naive in assuming I can simply subtract the values of one raw array from the other?

I am trying to recreate the

Submitted by rawhide on

I am trying to recreate the pixel statistics (min/max/avg/σ) that appear in RawDigger (per channel). Am I correct in assuming you are the authors of this software? If yes, then which LibRaw parameter was used to obtain these values?
Thanks

RawDigger calculated

Submitted by lexa on

RawDigger calculated statistics on unaltered RAW values, the only processing performed is black level subtraction (if not turned off by user).

-- Alex Tutubalin @LibRaw LLC

OK, but it's not possible to

Submitted by rawhide on

OK, but it's not possible to get the R/G/B/G2 values without postprocessing, or am I confused?

Quote from above:

Submitted by lexa on

Quote from above:

> You can use 4channels sample to split source RAW into four separate TIFF files.

(or you can use 4channels sample source as reference for your code)

-- Alex Tutubalin @LibRaw LLC

samples/unprocessed_raw.cpp

Submitted by lexa on

samples/unprocessed_raw.cpp sample is also a good starting point if you need unaltered raw values.

-- Alex Tutubalin @LibRaw LLC