Raw files made with N8 much faster loaded than RAF files made with X-H2. Why?

DxO PhotoLab Feature requests
1

As to read in the topic: to load nikon files into PL 8 is much faster than Fuji x-trans files. Is this my own problem or is it real?
Thanx for Ideas…

2

Fuji X sensors are more complicated to develop because of their X-Trans colour filter array pattern. Most other manufacturers’ sensors use a Bayer pattern. Read more about it here: Color filter array - Wikipedia

Moreover, DxO has added X-Trans development support (relatively) recently and I suppose that the respective code is not as advanced/optimised yet.

3

This is expected behavior with any software. Most cameras (Nikon, Canon, Sony, …) use standard 2x2 Bayer CFA pattern (RGGB). X-Trans CFA uses a specific 6x6 RGB pattern to minimize possibility of moire and similar artifacts. It comes at a price of demosaicking performance, since for (good quality) X-Trans demosaicking more neighbouring pixels have to be taken into account than for Bayer. See e.g. the following links (they are old but main points still hold).

https://www.libraw.org/node/2677

https://medium.com/@nevermindhim/x-trans-vs-bayer-fantastic-claims-and-how-to-test-them-475b4f1b7fae

BTW, most studies on denoising algorithms deal with Bayer sensors, so processing X-Trans files is a bit behind also in this respect.

4

Thanks for your interest, the different algorithms Bayer/X-trans are well known, but I don’t speak about development, denoising and so on. Only the loading of the files with no allowed action is significant different. And, by the way, the nikon files are larger than the fuji files. OK, no need to howl in sadness :smiley:…

5

Contradiction.

When opening a raw file PL parses metadata (libraw open_file(), probably modified), the unpacks the image into memory (libraw unpack()) and then uses its own code to demosaic the raw data. This last step takes majority of time. Only then it can display the preview. Execution of open_file() and unpack() functions depends on things like compression used, amount and encoding of metadata, so not too much on CFA type. However demosaicking takes several times more time, and it does depend on CFA.

6

Ok, I agree with this explanation. Thank you!