How to measure dynamic range of bracketed shots

[feppe]

One for the tech experts here: how can I measure the dynamic range (in stops) of a scene? I have a series of nighttime cityscape bracked shots, and would like to know how much dynamic range it contained originally.

I'm not concerned with exact technicalities with noise floors and tonal ranges, just a rough measure to see whether the same scene could be captured with slide, negative and/or B&W film. Yes, film.

Below example from a bracketed panorama, 5, 20 and 80 second exposures.

[bjanes]

One for the tech experts here: how can I measure the dynamic range (in stops) of a scene? I have a series of nighttime cityscape bracked shots, and would like to know how much dynamic range it contained originally.

I'm not concerned with exact technicalities with noise floors and tonal ranges, just a rough measure to see whether the same scene could be captured with slide, negative and/or B&W film. Yes, film.

Below example from a bracketed panorama, 5, 20 and 80 second exposures.

You can determine the photographic dynamic range by determining the luminance of the brightest area of the image demonstrating useful detail and the luminance of the darkest image area demonstrating useful detail. To do this you can't use the pixel value in the converted file since it has undergone compression during the rendering process and is gamma encoded. Furthermore, Photoshop will not do an HDR merge on your images since it thinks they were taken by different cameras.

To determine the scene luminances you need scene referred images. Raw files are scene referred, so you could look at the pixel values of the brightest and darkest areas as described above. If you can get meaningful averages from any one raw file, you could use those data. Otherwise, you would have to use two or more images and figure the offsets according to the exposure. To look at raw data you could use Rawnalize, Iris, or DCRaw. If you want to use ACR and Photoshop, you could use the methods described on the ICC Web Site.

[Guillermo Luijk]

A good way to estimate the luminance DR of your scene using the RAW file(s) is to assume sensor linearity (of course this is an approximation, but very valid on most cameras), take the least exposed RAW file, and extract its RAW data with:

dcraw -v -d -r 1 1 1 1 -4 -T -t 0 file.cr2

Open the resulting TIFF in Histogrammar (set gamma to 1 if another value was selected), activate RAW mode (choose the proper Bayer topology), and look at the log histogram in stops to find out along how many EV the information spreads. Since in high DR situations the deep shadows will be poorly represented because of the lack of levels and noise, it is recommended to apply the same procedure to the most exposed RAW files to obtain better knowledge of the DR in the shadows.

Example: this scene was captured with two shots 4 EV apart:



The RAW histograms produced (top least exposed RAW file, middle most exposed RAW file, bottom optimum RAW blend of the former files):



Looking at the histogram we can quantify the DR of this scene in about 12EV.

Regards

PS: I wish Gabor implements someday this log scale in Rawnalyze's histograms.