Processing Olympus Hi-Resolution mode

I process my photos with a combination of dcraw, imagemagick and g’mic but only just discovered this site!

I see a thread on Pentax pixel-shift, but does anyone have any insight into Olympus Hi-Res mode? I don’t have the hardware but RAW and JPG files are available at Imaging Resorce. I looked at the 80Mp RAW and corresponding 50Mp JPG files from the E-M1 MkII.

Dcraw extracts an 80Mp image from the RAW, but it looks soft and I am unable to process it to anywhere near the sharpness of the 50Mb out-of-camera JPEG. I tried RT on the same file - it loads and displays an image but when I try to save it appears to work, but nothing is written to the disk!

It is also a mystery to me how (a) Olympus creates an 80Mp from 8 x 20Mp images and (b) why and how they downscale to 50Mb for the JPEG output.

In dcraw the -s switch has no effect - it appears there is only one 80Mp image in there. Are they doing some kind of tricky demosaic with multiple images?

RT 4.2.1241 on Debian / dcraw 9.2.

1 Like

Olympus’ HiRes raws are not multiframe files, they are single frame coming from a mysterious incamera mix of 8 shots. Half of these shots have been captured at a shift of 0.7X diagonal with a lot of overlaping (about equal to 2X stronger than usual aa filter).

From these 8 shots a full color (3 colors per pixel) 40Mp diagonal grid are then mixed and then this gets transformed to a Bayer 80Mp grid (1 color per pixel) … there is a slight data loss as R and B data loose 1/2 of their sampling density i.e we have 20Mp R B instead of the 40Mp captured.

Because of the overlaping we take a soft picture (but this also gives perfect antialiasing ) … so for the Hires raws we need to use 1.5-2.0X stronger sharpening … compared with a normal raw coming from a sensor with no aa filter we need to increase radius by 1.4 and strength about so, to take about equal per pixel aquity (and a lot better details).

2 Likes

Thanks for the quick reply! That is very interesting information.

You are right - I was much too cautious in sharpening the output. Now it looks very good.

Perhaps the E-M1 MkII can position the sensor more accurately than its predecessors and we will see a bit more detail from the images. It looks like quite a jump compared to the standard 20Mp capture. More than I recall from the E-M5 MkII samples.