Sorry, this comparison could be meaningless, because we may be reviewing just the effect of the (unrevealed) image resizing method on the resolution.
- If the K-3's image was shrunk using the Nearest Neighbor method, it would have been the sharper image.
- If the K-3's image was shrunk by any other method (which I assume because Bicubic is default), it was the less sharp one.
Since the 2nd picture is less sharp by a big margin, I assume that that's the one which has been shrunk, which in turn implies it came from the K-3.
If the "test" was done the other way round, i.e. if the K-5iis' image was inflated to 24mp to match the K-3, then the sharpness difference would have been just the other way round.
Secondly (probably mentioned already), the AA simulation mode of the K-3 is unrevealed.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[edit] BUT THEN: the Exif data of the second (the unsharp) image reveals a sensor resolution which matches the K-5, not the K-3. Which could imply one out of the following:
- The K-3 Image was shrunk using the Nearest Neigbor method (w/o telling us)
. - [conspiration theory] This is a trap, i.e. false sensor resolution information has been injected into the second file using e.g. ExifTool. Suspcious: the first file doesn't have this tag at all. Because I don't assume that Ricoh will have removed this tag between K-5 and K-3, this suggests a manual change to the metadata to the conspiration theorist
. - Does photoshop tweak the sensor resolution tag, along with a resolution change, so that it doesn't mean anything?
. - All cameras after the K20D, i.e. the K-7 and K-5... are vulnerable to shutter blur, according to Falk Lumo's research. Do we see shutter blur of the K-5...? Maybe, this vulnerability has been removed from the K-3 then.