Originally posted by Oakland Rob You can look at the results of people running handheld PS shots through PDCU here; it removes a LOT of motion. What I don't recall is if that just means it effectively just uses one frame of the PS shot, or whether it does some masking a la deghosting when you do HDRs. The key to tell would be whether the static parts of the shot have increased detail.
So I'm kinda wondering if the PS II onboard thing is just basically a software improvement in de-motioning and stacking, combined with a faster processor to get it done. Basically a Photoshop update along with a new laptop in your camera. Which begs the question of whether one should upgrade the camera, or just spend upgrade Rawtherapee and get a faster processor on your computer and do the motion work in post, which allows for you to work with RAW and to have more control over the process. But maybe it's more than that—maybe the actual capture is improved. I'd sure like to know the details before I bought a K-1ii, since that's the key feature.
Ricoh's description of the system seems to imply that it is much more than just Rawtherapee on a chip. They are using SR system data in ways not used on the K-1 and not available to Rawtherapee.
I doubt the K-1 in pixelshift mode puts out all the SR system data that the K-1ii is using to do dynamic pixelshift. At the very least, the K-1 is unlikely to store SR system data with the pixelshift Raws because SR is disabled on the K-1 during pixelshift. It's also possible (and seems likely to me) that the SR system sensors on the mainboard have been upgraded for the K-1ii to more accurately measure camera motions over the 0.75+ seconds of a pixelshift exposure set. That suggests that the K-1iI can do things that the K-1 + Rawtherapee cannot.
This issue of in-camera processing versus post-processing is very interesting and does not have any obvious answers. On the one hand, post processing can use much more sophisticated and time-consuming algorithms running on a much more powerful computing platform. But, on the other hand, in-camera processing has much more access to internal data about the sensor, system conditions, lens, and SR sensor measurements. The issue also affects whether post-processing can replicate the noise-surpression offered by the accelerator unit. Whether brute force post-process computing can replicate the benefits of more internal data is not obvious.