Originally posted by jpzk One thing that I don't understand is how you "interpolated" the image to 26Mb in ACR. I do realize that you's therefore decrease the grain "size" and that would make, possibly (?) make the software now recognize it as "real noise", grain being not so large.
When you open a file in Camera Raw, there is an option at the bottom to select image resolution. My theory is/was that if you raise the resolution, that ACR will effectively reduce the pixel size which in turn would allow the noise reduction software to treat the image differently.
At this point, I can't honestly say if it improved things much, but it seems to of changed final output in some ways.
On a same note, I've noticed the Kx shares many of the same attributes as the K-5 when it comes to high ISO noise patterns. In the sense that it too presents a sort of ceiling to which NR management begins to loose it's effect on noise patterns in high ISO images. I suspect this is all relative to the way the NR software filters or identifies noise in an image and at which point, it noise becomes to prevalent, the software can no longer distinguish between noise and that of veritable image data.
Though one thing I've noticed is where this type of noise(Kx and K-5) is quite consistent or uniform. Which may actually end-up being a good thing as may allow for a specific filtering systems to better target and separate noise from actual image detail. In which case, if you look over the ISO51200 RAW image carefully, you'll notice that there is a good deal of detail within the RAW image despite the noise patterns. Which tells me that we may see yet more major advancements in noise reduction in the near future.
To which I'd say... who knows how far this could go. ISO102400 may be a reality much sooner than we though