Originally posted by reh321 I am quite sure that Sony is also has 'processing' in the path to their 'raw' files - it may be integrated into their sensor as another layer, but it is there somewhere.
It has never been established that there are any cameras with no NR applied in their high ISO images. People are complaining about what is standard practice in the industry. The whole industry is wrong in it's direction, they are right. One could argue, that's why they aren't camera company execs. This whole concept of "baked photos" is a part of a "artifacts make good images" movement. Personally for myself, the scene I'm photographing is good, artifacts are bad. Others it would seem have an opposing opinion.
---------- Post added 06-02-18 at 10:21 AM ----------
Originally posted by BigMackCam And now, here's the good news. I've been working with the ISO 6400 K-1II image in Darktable for a couple of hours, and it seems to respond rather well to additional processing, such that detail can actually be recovered (or, more accurately, emphasised - as it doesn't seem to have been lost in the first place).
Something many don't understand. There is a lot of detail in every image that is lost because contrast values don't permit it. You can have millions of gradations in the white, the high end of a digital file, but the human eye can't distinguish a million colours in the high end of the curve. By lopping off the whole bottom end just using the high end and contrasting the resulting file, you can bring out all kinds of detail. When you have millions of colours, a digital 16 value difference is imperceptible. But adjust your curve to the black end, where in jpeg, ou only have 16 values every value is perceptible. Nothing is more frustrating than try to bring out high end values without spot tools. If you get the detail you want in the high end, your dark end is completely black.
So the biggest factor i want to know in these kinds of comparisons is "is the centre of the histogram curve in both images the same?" Because moving the centre of the curve effects high and low end contrast. If the curve has been shifted to the right by your importing software, the left side fo the curve has more gradients to portray the low end end, shifted to the left gives you more gradient values in the high end.
Hence, a minor shift in the centre point of the curve will produce different contrast values in different parts of the picture and bring out detail in one end while suppressing it on the other. Most of the time we want good contrast values on the subject and we don't care about the rest. (especially in high tone photographs).and will create the illusion of more detail in one end or the other, just by making captured detail visible. But the detail was always there. This is what is so difficult for people when looking at these images. Differences in the adjustment of the centre point in levels causes different levels of detail in different parts of the photographs. Up until now it didn't matter because there were differences in resolution between different sensors that were so noticeable that such things didn't matter.
The comparison here has gotten so fine, that every little thing matters, to the point that many previously ignored factors are now starting to produce very uneven results. I suspect that some of the determinations people are trying to make can really only be examined using Byte value comparisons of the files under examination as a series of bytes, not a photograph. And I know of no software that can do that. The processes being used are just not good enough for the kinds of results being claimed.
It's pretty obvious to me what's happening. The measuring tools are obsolete. Leaving the door open for the haters at DPR to just make up stuff and spout opinion as if it has something to do with some really inadequate (for the task) example photos. We are trying to measure quark level differences with a tape measure.
To summarize, the level of difference in the original file is less than the values that can be changed by contrast, setting levels controls, NR and many other factors which all taken together present an insurmountable number of uncontrollable variables, that make more difference to the final output than the characteristics of the original files do..