Originally posted by falconeye I just had a closer look at the samples JohnBee uploaded. And I noticed a strange thing, unrelated to the AA filter thing and causing a sharpening artefact which made me take the closer look.
The K-5 pixels almost come in in 2x2 blocks!! There is only minimal variation within a block. A demosaicing routine should never produce such results. I attach a second 800% crop to showcase this. What looks like a pixel are 2x2 pixels. Just look more closely!
Falk, perhaps it would be better if you worked with the originals?
D7000, ISO100 RAW/NEF Sample; http://216.18.212.226/PRODS/D7000/FULLRES/D7000hSLI00100_NR0.NEF K-5, ISO100 RAW/DNG Sample; http://216.18.212.226/PRODS/K5/FULLRES/K5hSLI00100_NR_0.DNG
Let me know if this helps.
---------- Post added 11-18-10 at 03:21 PM ----------
Originally posted by Urkeldaedalus Of all of the people who own the K-5, not one person who's reviewed it so far that I've seen has talked about soft photos; in fact, the camera has been praised for its image quality up to this point. Now Dpreview and Image Resource do some admittedly flawed "tests" and all of a sudden the K-5 has soft photos compared to the competition. I know why they do these comparisons and I can see some value in it, but I feel like it's a hard way to actually judge cameras against one another. You're really not seeing any of these cameras at their best.
I believe both DPReview and Imaging Resources have redone their testing in an effort to correct any discrepancies.
However, I think in a case such as this one, we have a little more at our disposition than mere personal observation.
Taking for example, that Image Resource offers controlled testing with the same lens and settings from other Pentax camera's(ie, the K-7). Which makes it more likely to identify any/if discrepancies in the process.
Which raises the question.
Based on the K-5 specs, wouldn't it be fair to expect the K-5 to either match or exceed the K-7 in nominal conditions?
To which, I would say, yes it is.
Unfortunately, at this time, it looks as though the K-7 samples from Imaging Resources have better IQ than the K-5.
And though this may be due to flawed testing(I know IR mentioned this). However, since the results seem consistent with other testing sites also, I think it gives use good cause for investigation.