2
\$\begingroup\$

I've just read Karl Lang's excellent whitepaper - Rendering the Print: the Art of Photography. In it, he mentions that the ratio of light to dark in a 'natural scene' can have a ratio of 100,000:1, a modern high-end sensor can only capture a range of 8,000 to 1, while a modern pigment printer can only print something like 250:1. He also mentions that it is possible to capture the full dynamic range of a scene using a spectral pixmap though he says this is not practical for general use for reasons of cost and duration of capture. This got me wondering about RAW data generated from sources other than consumer cameras, or created artificially to represent the full possible dynamic range. Do any such RAW files exist? I appreciate that shooting HDR is a way to achieve a similar end, but I'm interested in seeing (and hopefully playing with) a RAW file containing data for the full potential dynamic range.

Improve this question
\$\endgroup\$
9
  • \$\begingroup\$ "a modern high-end sensor can only capture a range of 8,000 to 1" -- Is that not the answer to your question? That's what a Raw image is: the sensor data. \$\endgroup\$
    – user4894
    Nov 20, 2015 at 4:09
  • 1
    \$\begingroup\$ Just because a natural scene "can" have a ratio of 100,000:1 doesn't mean all natural scenes "do" have a ratio of 100,000:1. Can you explain in your question exactly what he means by "natural scene"? \$\endgroup\$
    – Michael C
    Nov 20, 2015 at 4:14
  • \$\begingroup\$ @MichaelClark What phrase would you prefer to mean a scene with maximal dynamic range found in natural surroundings? I am just using the same semantics as Karl Lang uses in the whitepaper I referenced. I don't think at any point in my question I stated that all natural scenes have a ratio of 100,000:1. In face I specifically used the word 'can' which is like saying 'it is possible for'. I think it's pretty obvious that the ratio changes depending on time of day, environmental factors etc. \$\endgroup\$
    – Undistraction
    Nov 20, 2015 at 17:20
  • \$\begingroup\$ @user4894 Obviously that is what a RAW image is, but you make a false assumption - that the only sensor that can generate RAW data is a high-end commercial sensor. See the comment above yours for an example. \$\endgroup\$
    – Undistraction
    Nov 20, 2015 at 17:30
  • 1
    \$\begingroup\$ @MichaelClark I agree my use of the word image is erroneous. I will change it to data in the question text. \$\endgroup\$
    – Undistraction
    Nov 20, 2015 at 17:59

3 Answers 3

Reset to default
6
\$\begingroup\$

To achieve 100,000:1 dynamic range the raw file would need to store 17 or more bits per pixel, since 2^16 =65536. Which is by no means normal but could be done with CG data.

Improve this answer
\$\endgroup\$
1
  • 2
    \$\begingroup\$ True, and you're keeping the sensor design essentially the same, and altering the ADC resolution. Alternatively, you could have two wells represent a single pixel, one standard, one overflow. Or in another design, you could reset the well once it hits saturation and record the count of of saturation events . That way you can have a pixel represented by multiple 16 bit (or smaller) values in a single RAW file as the op seems to want. \$\endgroup\$
    – user77983
    Jan 17, 2019 at 21:56
4
\$\begingroup\$

In darktable, at least, there exists the ability to create an HDR "RAW" file (DNG format) by merging two or more bracketed exposure RAW files, which can then be processed as any other RAW file. It's kind of an oddball feature, and very sensitive to any movement in the frame between shots compared to other forms of exposure merging (even slightly drifting clouds can create bad artifacts), but it does make for a nice HDR workflow when it works. Some implementation details (source):

first, a disclaimer: this merging process is really dumb. it was very fast to implement, that's why we have it. you will probably run into situations where it's not appropriate to use it.

it just merges the plain raw pixels, the only thing done to them is subtraction of black level, scaling to white according to their exposure/aperture/iso settings, and then the values are written to dng.

doing it that way doesn't require to extract a `camera response curve' as needed for jpg. in fact, it doesn't even demosaic.

UPDATE: I recently became aware of HDRMerge, but haven't used it yet. It appears to be similar to the darktable approach of combining the RAW data directly and outputting a DNG file with otherwise unprocessed data.

Improve this answer
\$\endgroup\$
0
\$\begingroup\$

You can download raw Hubble image data here in FITS format, most of this data has never been processed into viewable images. This tutorial explains how you can process these files using the free of charge program FITS Liberator to e.g. 16 bit tiff files that you can then process further using other image processing software.

Improve this answer
\$\endgroup\$
1
  • 1
    \$\begingroup\$ The implication is that these are files representing scenes with very high dynamic range. Is that the case with these images? Maybe you could give a bit more detail on that. \$\endgroup\$
    – mattdm
    Nov 21, 2015 at 21:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.