I have already posted this on LL but didn't seem of much interest to the audience. However I consider it very interesting.
All of us know that captured DR can be expanded by shooting more than once at different expsoure values, and properly blending the information obtained from all those shots (in general highlights from the least exposed shots, and noiseless shadows from the most exposed ones).
Unluckily most programs performing this task (Photomatix, HDR in PS, Enfuse) do their proper RAW development and tone mapping, so the user looses control over the process.
What I tried to achieve is to fuse several RAW files, but obtaining another RAW file as output. This resulting RAW file is noise free (as much as the source files allow for) and preserves the highlights, so looking at the information contained on it it can be considered an HDR RAW file. But the new thing is that now the user can develop it in his own RAW developer or preferred tone mapping software.
I.e., I have performed an optimum (maximum noise reduction and maximum sharpness preservation via non-progressive blending) mix of N RAW files into a resulting RAW file (16-bit DNG actually) that summarizes all the useful information found in the original files. Now it is up to the user what to do next (develop it with ACR, tone map it with Photomatix,...).
What do you think?
The original so as the resulting DNG are ready for download in the article: ZERO NOISE VIRTUAL RAW.
The scene was captured through 2 exposures 4 stops apart. When comparing the resulting DNG to the original files, we obtain:
With proper RAW development adjustments in ACR we get straight from it this tone mapped HDR picture:
BR