[Robert N]

For new owners of the K3-iii, it would be interesting to know how the pixel shift performs. Points of interest are for when shooting in RAW:

As a rough percentage, how much is the resolution increased (I believe the file size is the same as for a single shot).
How does the resolution compare to a single frame of the K1 (for owners of both the K-3iii & K1).
Is the pixel shift effective at compensating for moving objects (typically trees blowing in the wind).

Looking forward to your replies.

[house]

I'ts quite tricky to evaluate pixelshift. Some scenes show barely detectable improvements whilst others show significant improvements. Atmospheric haze etc will cut away a lot of the benefit as will diffraction. It sounds odd but I also feel that some scenes don't have detail at the scale where pixelshift is active.

[madbrain]

I can seldom tell the difference between PS and non PS images on my K1 II. Probably because I'm not shooting with the sharpest lenses.

[Rondec]

I can seldom tell the difference between PS and non PS images on my K1 II. Probably because I'm not shooting with the sharpest lenses.

I wouldn't think the K-3 III would be any better at pixel shift than previous cameras -- except maybe if you are just shooting jpegs with motion correction enabled -- it might be better then.

The difference in PS and non PS images is seen in post processing. But you typically have to send the image through either DCU or Raw Therapee to correct for motion, then you can bump shadows and sharpen a pixel shift image more without artifacts showing up than a single shot image. Without those steps the images really will appear the same.

[Ishpuini]

I use Lightroom for PP, and I remember that we could only see some effect in there once Lightroom explicitely supported pixel shifted RAWs out of my K-3II. Without that update, Lightroom just took the first exposure out of the PS file, so obviously no difference was visible. I guess the same would be required now, and with other RAW converters as well? So best wait until the proper RAW developer release.

[madbrain]

I use lightroom for PP as well. It's ok if shooting PS on tripod.

[house]

you can bump shadows and sharpen a pixel shift image more without artifacts showing up

yup that's certainly the case. Withouth the extra sharpening the difference is very subtle. Except on some cases with lots of detail at pixel level where it's immediately noticeable.

[Rondec]

I use lightroom for PP as well. It's ok if shooting PS on tripod.

Depends on what you are shooting. lightroom doesn't do motion correction so if there is any motion in the scene you will get artifacts. I particularly notice this with leaves on trees and flowing water. I do always shot from a tripod -- except for the dynamic pixel shift, hand held is useless in these situations.

Raw Therapee has the ability to show the motion mask and it is surprising how, even on a still day, there is often motion in foliage.

---------- Post added 05-18-21 at 08:39 AM ----------

yup that's certainly the case. Withouth the extra sharpening the difference is very subtle. Except on some cases with lots of detail at pixel level where it's immediately noticeable.

I think the other place that we haven't mentioned is in eliminating moiré.

[xandos]

It seems to me that differences in colour fidelity should always exist, but in many cases where the sharpness is limited by the optics rather than by the sensor, pixel shift shouldn't have much of an effect on sharpness. In those cases where the lens limits the sharpness, moire should also not show up.

[clackers]

As a rough percentage, how much is the resolution increased

Zero increase, Robert!

26Mp in a pixel shift image, just the same as with it switched off!

But each pixel has had an additional three checks on the accuracy of its contents.

There will be noise reduction, and truer edges to objects instead of smears. As Rondec points out, you can then push the sharpness sliders in Lightroom or whatever further without being alarmed.

[Rico]

What is the shot to shot time with the K-3III using Pixel-Shift? It takes the K-1 5 seconds until the camera is ready due to the slow UHS-I bus. The UHS-II bus on the K-3III should reduce the time till the camera is ready to fire again.

Pixel-shift definitely delivers superior true color fidelity. It's so good there is no real need for using a color checker passport to get accurate color.

Clackers is right there is no increase in resolution. There is better noise control at higher ISO's and a more 3D real world atmospheric space to the files.

[wkraus]

What is the shot to shot time with the K-3III using Pixel-Shift? It takes the K-1 5 seconds until the camera is ready due to the slow UHS-I bus. The UHS-II bus on the K-3III should reduce the time till the camera is ready to fire again.

Pixel-shift definitely delivers superior true color fidelity. It's so good there is no real need for using a color checker passport to get accurate color.

Clackers is right there is no increase in resolution. There is better noise control at higher ISO's and a more 3D real world atmospheric space to the files.

Are you sure about the 5 seconds? This does not seem to be my experience with the K-1 (although I never measured the time it needed). I just did a quick test with the K-3 III with an exposure time of 1/125 and it was able to fire again after 8.5 seconds using a V90 card.

The resolution thing is not so simple. The amount of pixels is obviously the same, but the ability to reproduce small details accurately is quite a bit higher with Pixelshift.

Simplifying some of the problems involved, sampling theory says that for the green channel the resolution of a Bayer sensor is the same as that of a monochrome sensor; for the red and blue channels it is half of that. For the usable resolution the demosaicing technology is critical. This article concludes that “adaptive interpolation can lead to X and Y axis resolution as high as a monochrome array with the same pixel pitch” (p. 245). Practically, however, much of the information near the resolution limit risks being destroyed by aliasing and demosaicing artifacts. Part of the beauty of Pixelshift is that false colours resulting from these ugly things completely go away.


I would like to add that this holds for the Pentax implementation but less so for Sony’s. Judging from the A7R IV Pixelshift sample in DPR’s studio shot comparison tool, its output is not free of colour artifacts (but four times the pixel size).

[andy271]

pixel shift pentax k-1 RawTherapee v5.8 is the best solution ive even compared to my gfx 100s and haselblad 907x and the pixel shift from the k-1 is better i must point out im using the best lenses available for each system

[stevebrot]

But each pixel has had an additional three checks on the accuracy of its contents.

No more disappearing small flowers either!


Steve

[wkraus]

As a follow-up to my last post, here is a test shot (an old Paterson test target) developed with and without Pixelshift to demonstrate its effects. The lens used is the SMC D FA 50mm F2.8 Macro at f/6.3. I used three different raw converters, Adobe Camera Raw 12.2 (no Pixelshift functionality as yet), RawTherapee 5.8, and Digital Camera Utility 5.9. Incidentally, this also demonstrates how differently the demosaicing works at pixel level – but the test pattern is a real stress test, under normal conditions the differences are less obvious.

These are full-size crops. I edited the images for identical luminosity and contrast and neutral grey, but did not attempt to get similar colours. Due to the complexity involved, I also did not try to apply the same sharpening. In DCU sharpening is set to zero but nevertheless the software seems to apply some rather crude sharpening.

Adone Camera Raw:


RawTherapee:




Digital Camera Utility:




Feel free to download the images if you want to take a closer look.