Originally posted by Kunzite In the Imaging Resource interview, Mr. Takashi Arai talks about a kind of processing which cannot be reproduced when post processing the RAW files:
Well, he is trying to sell a product and I don't blame him for that.
The actual claim he made is non-sensical (see below) and possibly owed to a translation issue.
We cannot be sure what he really said (unless there is an original transcript).
Originally posted by Kunzite and then an accessory unit does a kind of signal processing which cannot be obtained by just software processing mechanism without degrading the resolving performance of the sensor."
Anything that is computable at all can be achieved through software processing. There is nothing hardware can achieve that software couldn't achieve given access to the same data. If there is an actual claim regarding the inability of software to achieve the same effect outside the camera, it would have be to be argued differently.
Everything about the K-1 II files points to a standard ex post facto processing. This, combined with the fact that modern Sony sensors are essentially black boxes which spit out the final data without offering options to interfere at the low-level A/D conversion level, make any claims regarding a processing that could not be replicated outside the camera extremely unlikely to be true.
If someone were so inclined, I'm rather certain they could replicate the look of a K-1 II image from a K-1 image. The demonstrated attenuation of spatial high frequencies is entirely consistent with regular post capture denoising. The merit of the "accelerator" is not that it does something unique, but that it relieves the user from doing it.
Originally posted by Kunzite By the way, here's a good, detailed explanation why the accelerator isn't optional:
I've read it and I agree with clackers that there is no explanation as to why it wasn't made optional.
It seems there is a hint as to why the processing starts so unusually early (ISO 640). If they had let the denoising kick in at ISO 3200, say, it would have been very obvious that something odd is going on as the look of the images would dramatically change from one ISO setting to the other. By gradually introducing more and more processing, starting very early, the processing becomes rather inconspicuous. It fooled DPReview the first time around (-> KP).
I speculate that reasons that there hasn't been a firmware update that makes the processing optional could include Japanese pride that shies away from implicitly admitting mistakes, and the fact that treating the processing as non-optional gives credence to the idea that the K-1 II's low noise levels are not just the result of an image processing algorithm (running on specialised hardware, the so-called "accelerator").