Pentax K-3II pixel shift samples show massive clarity & resultion enhancement at same pixelcount

[pixelsmithy]

The most interesting comparison now will be what the same subject looks like when shot with a foveon.

That's the most related technology and it is known for very good detail.

I'm sure that someone will do it at some point, simply because it would make an interesting article. When they do, I'm sure they will also compare the Pentax pixel-shift resolution against the Foveon at higher ISOs (where the Foveon will certainly underperform). Even a lot of Foveon owners are frustrated with Foveon (if you read their forum). There are always limitations and trade-offs in camera design, but the K-3 II appears to be the best COMBINATION of features in one package EVER.
--

Check out my (UPDATED 05-19-2015) Gallery! http://www.dpreview.com/galleries/355548730/photos/slideshow

 

[geo444]

.

this P E N T A X is F A N T A S T I C !

Now when you need it, you have a ~39 MPix equivalent sensor, but at 24 MPix filesize !

and... this without the drawbacks of the Foveon sensor tech :

- poor Color Separation : a Photon can emit an Electron in each Color Layer... or none...

= very low QE ~ 20-25 % (~NiKon D2x)



poor Color Separation : a Photon can emit an Electron in each Color Layer... or none...



--
http://en.astronomike.net/m/6809.html
http://en.astronomike.net/m/134251.html

 

[Wallace Ross]

In the urban image there are too many artifacts. The people walking the trees moving and the cars of course. But in the forest image it looks very good. It looks like the use for this will be somewhat limited as Ricoh has suggested. I am still amazed at the original K-3 for detail so this if used properly could be great. P.S. there seems to be a stuck pixel(s) in the forest image on the right side near the base of the tree.

 

[Roland Karlsson]

In the urban image there are too many artifacts. The people walking the trees moving and the cars of course. But in the forest image it looks very good. It looks like the use for this will be somewhat limited as Ricoh has suggested. I am still amazed at the original K-3 for detail so this if used properly could be great. P.S. there seems to be a stuck pixel(s) in the forest image on the right side near the base of the tree.

With more mature software for this functionality, this problem might be eliminated. Other stacking software use motion detection and fallback methods to treat moving parts. So can pixel shift do.

 

[falconeyes]

Motion detection isn't always possible for this pixel shift method.

First, you'd need to demosaice all 4 shift images first to analyze them for change and possible motion. A huge hit on performance but doable for sure.

But more importantly, the worst artefacts come from subtle movements of about 1 pixels between captures. And those wouldn't be recognizable because you cannot distinguish movement from color change. If you could, you'd not need to pixel shift in the first place.

So no, don't expect motion detection for pixel shift raw combination software which is meant for static subjects only.

 

[Roland Karlsson]

Motion detection isn't always possible for this pixel shift method.

First, you'd need to demosaice all 4 shift images first to analyze them for change and possible motion. A huge hit on performance but doable for sure.

But more importantly, the worst artefacts come from subtle movements of about 1 pixels between captures. And those wouldn't be recognizable because you cannot distinguish movement from color change. If you could, you'd not need to pixel shift in the first place.

So no, don't expect motion detection for pixel shift raw combination software which is meant for static subjects only.

Maybe you could do the conversion, assuming there is no motion, and then do detection (with FFT maybe) trying to find areas where there are obvious stripes or striped edges.

 

[kitsios_spyros]

I saw the thread from my mobile and I was waiting for the time to see the full size samples from my monitor at home. Wow! Very impressive quality. Alasing errors almost minimized, clarity is fantastic, almost foveon like, and there is a clear step up from the 16MP sensor.

There are some problems but I am really optimistic that next incarnations of this feature and software will decrease them. But especially for the scenes that one would anyhow use a tripod, this thing is a just WONDERFULL!! We have been discussing around two years that the 24MP sensor requires higher SS vs low MP sensors, so for critical scenes a tripod or higher SS (=>better SR?) was essenstial. This feature delivers. Plus that the new SR module increases the hand-holding stability to 4.5 stop and hopefully there will not be problems for BIF applications.

Well done Pentax!