Raw Therapee now CR3 ready ( but metadata is not supported yet) & lens blur correction

Feb 7, 2020
1
0
From the web site:

  • Automatically recover detail lost to lens blur (diffraction) using the new Capture Sharpening tool, located in the “Raw” tab. It takes place right after demosaicing, and as it works in linear space it is not prone to haloing. Capture Sharpening in combination with Post-Resize Sharpening allows for detailed and crisp results.
  • CR3 support: image data is decoded so you can process your raw files, but metadata is not supported yet. If you have an ICC or DCP input profile for your CR3-producing camera, you will need to point RawTherapee to it manually (Color tab > Color Management > Input Profile > Custom).
  • Improvements of various camera models (new DCP dual-illuminant input profiles, raw crops, white levels, etc.), speedups and optimizations to various tools, better memory management, various bug fixes. See the git log for details.
 

AlanF

Canon 5DSR II
Aug 16, 2012
6,226
4,108
The so-called recovery of detail lost to lens blur (diffraction) has puzzled me for ages, but I think I now know what is going on. Perhaps the experts could correct me if I am wrong. Basically, diffraction causes loss of contrast as the Airy disks of blur get larger with decreasing aperture and when they overlap all fine detail that is closer than the radii of the disks is lost. Those algorithms used by DPP etc don't recover that lost detail, which are lost forever, but what they do is to clean up the noise in the detail that still remains in the details that are coarser than the diffraction limited but whose contrast is lowered by diffraction.
 

Joules

EOS 7D MK II
Jul 16, 2017
583
542
Hamburg, Germany
The so-called recovery of detail lost to lens blur (diffraction) has puzzled me for ages, but I think I now know what is going on. Perhaps the experts could correct me if I am wrong.
I think these techniques are just referring to applying some Form of blind deconvolution, which does reduce the effects of point spread in the image. This point spread is not just from Airy disks, the degree of lens correction, motion blur and disturbances in the atmosphere should also contribute to the kind of blur that is addressed by these filters. That they specifically mention ringing sounds very much like a blind deconvolution algorithm.

Under the presence of noise and no exact knowledge of the real point spread function, deconvolution does definitely not reconstruct the absolute truth a perfect imaging system could record. But it is also more than just sharpening by creating some fake details.

I find it really fascinating, but I'm not deep enough into it to know at which point blured detail is considered to be actually lost.