Support for Pentax pixel shift files #3489

Deconvolution sharpen ? more detail ? :wink:

I send for You link to orginal DNG 


There must be something wrong. I get completely different result (much worse) with your DNG


My compilation is best ? :wink:

You have AMD or Intel processor ?

AMD version last or Intel old be here :

I disable auto CA correction, You can use my pp3 file ? only change demosaic metod to pixel shift , color correction step to 0 ?

BR MK

ps. Compilation for Linux Mint 18 64-bit KDE :wink:

Left = pixelshift by the looks of it. Although in this comparison the sharpness of defined edges is only slightly improved to my eye, areas of more solid tone definitely look more smudged to me in the rightmost image.

For general/everyday use I tend to agree, but I think it does make a difference to details when printing large, and I believe it helps to avoid Moiré patterns and decrease noise in general, too.

That’s quite an improvement!


Here’s a pixelshift RAW back from when I first tested the feature shortly after getting the camera:

IMGP0088.DNG (Pentax K-3 II, Pentax HD DA 20-40mm at 36mm focal length, f/16, 1/3sec)

Unlike the previous example shot’s results above (boots), the following images did have additional sharpening (“fine sharpening” in the case of the pixelshift version) applied in DCU. I’d be interested to see what results you can get in RawTherapee on a single one of the four shots.

From original thread:

I used Pentax Digital Camera Utility (PDCU) 5.4 as outlined by the Pentax Forums K-3 II review here. No lens correction or other changes were applied. For the non-pixel shift shot, I used PDCU’s standard sharpening mode (the fine sharpening mode wasn’t helping in the same way it does with pixel shift shots and seemed to be having an undesirable effect), and applied as much of this standard sharpening as I could before angled edges got very obviously pixelly/jaggy.

2 Likes

Nice owl :slight_smile:
owl_05.png shows some sharpening halos in the both pictures but they are more pronounced in the right picture. I tried to match your version concerning details (obviously I failed) but without the halos. I didn’t try to match colour and exposure.
Again one of them is amaze demosaic and the other one pixelshift. I used same settings for both. Only exception is different demosaic method.
Btw: You’re right that in overkill example left one is pixelshift :slight_smile:

What I wrote there is nonsense. I must have been confused by the shifting :wink: In fact the coordinates of stuck pixels are the same for all frames. I ‘only’ have to move the current pixelshift processing to a different place in rt pipe. Then all the preprocessing could be enabled (to avoid confusion, it wasn’t disabled in gui, but bypassed in internal processing) again. Though I don’t know whether it will work fine for all preprocessing steps.

@heckflosse Ingo, I see a bit of magenta coloration on the black surface 
 which could mean that the Black Level used is a bit low 
 although I checked the exif where BL’s are are reported as all same (1024) 
there are pixelshift files where the BLs differ from frame to frame 
 does RT use the separate Black Levels of each frame ?

These look fine 
 but can you show what is happening at the 30-50 resolution trumpet ?

@Marek can you upload the results of DCU (motion on and off) on this file ( IMGP0505.PEF)

@ilias_giarimis Ilias, thanks for finding that. Currently RT does use the same values for all frames. I will see what I can do to use per frame black levels. Do you have a hint where I can find them?

What is the 30-50 resolution trumpet?

Edit: I think I know what you mean. Yes, I can show that in about 15 minutes.

Where I find this file for download ?

You can find it here. It’s the ISO 100 K1 pixelshift file.

@ilias_giarimis Top is pixelshift, bottom is amaze from first frame.

With exitool in --Pentax data I see four sessions (one for each frame) of tags 
 all four BlackPoint are at 0x0200 (all the four of them ) but I don’t understand how to come on each separately.
http://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/Pentax.html

In DNG format all the above also exist in subIFD 0xc61a

Motion off :

https://file.town/download/41qsy50tnuoc7tbfhc6pwrjd1

Motion on:

https://file.town/download/cthhy9ie2w5srbkxqzbluyrin

Amaze is better !! too strong crosshatch artefacts in pixelshift 
 as I remember, the same happened with RawDigger
DCU has much weaker artifacts in no motion mode (together with a general softening of details) 
 and almost no artefacts with motion on.

Hmm maybe this could help 
 http://lcav.epfl.ch/reproducible_research/VandewalleKAS07_1
https://infoscience.epfl.ch/record/89716/files/VandewalleKAS07.pdf 


Isn’t that preferable to the blue?

@ilias_giarimis Here’s a screenshot of pixelshift with False colour suppression steps set to 1

2 Likes

Ilias, I plan to do that. But in first step I would like to add the possibility to process all 3 frames of a Pentax HDR in hdrmerge. Unfortunately that does not work atm because the libraw version used in hdrmerge is not able to access the sub frames :frowning: