newbie problem with pixel-shift in RT 5.2

(David Wright) #1

I read that RT supports pixel-shift, so downloaded it. It’s going to take a while to figure out all the features, but for now I am just selecting ‘pixel-shift’ in the demosaicing method and saving the current image with that setting. This gives me an image that is sort of half-way pixel-shift. Something has be done, but it’s nothing like the PS output I get from DCU 5.6.

Here’s a slice of 3 versions of that image (Pentax K-70) at 400% - pixelshift from DCU, a single image extracted from the PEF file, and on the right, the output from RT. (Sorry-I couldn’t figure out how to embed an image):

pixel-shift from DCU on left, nonPS middle, RT pixel-shift right

Any thoughts on what I’m doing wrong?



@davidwright, welcome to the forum! To embed the images, just drag-and-drop or copy-and-paste.

(Mica) #3

Perhaps the pixelshoft play raw can be of some help: Play pixelshift

(David Wright) #4


Here it is:

(David Wright) #5

I looked at that, but it’s not really helpful.

Plus, the UI (and probably the features) are different. In that link they were using RT 5.0


(Ingo Weyrich) #6

Did you have a look at the motion mask in RT. Everything which is green when showing the motion mask, is demosaiced, everything else is pixelshift. The transitions between the green and non-green areas are smoothed (interpolated between demosaiced and pixelshift).

(David Wright) #7

OK, there is a lot of motion. But the reason I chose this image is the obvious color moire on that roof, which DCU was able to fix. This was image #31 out of my Pentax - I’ll try another pixel-shift image (of a still life) and see what happens.

On a (probably) unrelated note, the saved image out of RT, using the defaults (except for selecting Pixel Shift) are dark and flat, even tho the preview isn’t. This image above I had to increase the brightness by 40 and the contrast by 50 in PSE, to get what we see here.

But as I said, I’m a complete newbie with RT.


(Morgan Hardwood) #8

(Ingo Weyrich) #9

The colour moiré on the roof is almost completely removed also in RT version. But there are some small areas on the roof, which RT detects as motion. By setting Motion correction to off in RT you can verify this.

If you can share your raw file, I maybe can give you some hints.


(David Wright) #10

Thanks for those links.

So the fact that I don’t have a tone curve specific for the Pentax is the problem? And does that explain why the preview (in RT) looks different that the output file (insert at upper left)?

I noticed that when MS forced that latest update on everyone (the designers edition, or something like that, there were some subtle changes (for the worse) in colors.


(Morgan Hardwood) #11

@davidwright if the preview in RawTherapee’s Editor tab looks different than the saved image, and you’ve not used any of the tools marked as “1:1”, then that indicates that color management is not set up correctly on your system. I can’t tell with certainty without seeing your PP3 and without seeing your complete RT window (you cropped the screenshot unnecessarily).

(David Wright) #12

Thank you for your response. As I mentioned before, that was a very early image and I quickly realized that I needed a sturdier tripod, and I could not get away with just placing it on top of some dried grass, so my later images show less motion.

I went out this morning and shot my test scene - there is practically no motion detected in it (red circle around only motion):

Looking at various areas at 400%, it appears to me that RT is, in general, slightly worse than DCU in it’s default processing of Pixel Shift. But RT has a world of processing options; perhaps someone who masters those options could get a much better result. :slight_smile:

All of these comparsions are DCU left, RT center, single image (i.e. no PS) right

I will check if my dropbox accout is still active and place the rawfile there.


(David Wright) #13

Now I Notice that the Working Profile is ProPhoto.

How do I open the .pp3 file?

(Morgan Hardwood) #14

The working profile is not the problem.

You’re not using a monitor profile in RawTherapee. Are you using one in Windows or in your external image viewer?

You can save the PP3 file, rename the extension to .txt and upload it here, or just keep it as .pp3 and upload it using

Upload the raw file(s) too.

(Ingo Weyrich) #15

Imho your new example clearly shows that the center image has more fine detail than the left version.

(David Wright) #16

here they are

(Morgan Hardwood) #17

@davidwright the bin is empty.

(Andrew) #18

hi Ingo, I don’t have a Pentax and am not likely to have one sadly, but I’m curious about this. Referring to the centre “matrix”-ish section above the yellow one -

  • there’s a purple bit in the centre of the RT one which is not there on the left (on the boundary between light and dark)
  • there’s a lot more colour in the RT one, with a lot of green squares, and some magenta ones I think. The squares in the left section seem more uniform.

@davidwright, is the subject uniform in tone even up-close, please?

(Ingo Weyrich) #19

@RawConvert Have a look at the + like artifacts in PDCU version:

(Andrew) #20

I hadn’t noticed those but now you point them out, they’re striking. It’s as if PDCU is a bit random on brightness but uniform on colour, whereas RT appears the other way round.