Most underestimated camera currently available

Roland Karlsson wrote:

Alex Sarbu wrote:

Roland Karlsson wrote:

Alex Sarbu wrote:

Roland Karlsson wrote:

The two question regarding this noise suppression in the RAW data are

1. Is this noise suppression superior to one made in post process?
2. Does this noise suppression do anything degrading to the image?

Good questions. I'll add:

3. Are you measuring an actual image?

I am not measuring anything at all. This is from www.photonstophotos.net. And they are only measuring dynamic range and low level limit. No images involved.

You, them; the question stands.

It is actually the most important question in any performance evaluation. In this case, the answer is no, they're not measuring a real life image.

Whether there actually is any usable image or not is not measured. The noise reduction might decrease the resolution, maybe. That is not seen in the graphs.

Or maybe the noise reduction they're detecting is content aware thus their evaluation is invalid.

That is why I have my two questions.

But, I assume that you had a rhetorical question, and already knew that.

Alex

I am sorry Alex, but their measurements are probably just fine.

Maybe I wasn't clear - yes, their measurements are probably fine but what are they measuring?

I've seen people looking at such charts then crying, "aggressive noise reduction obliterates detail from ISO640!" - which is nonsense. There was no detail obliterated or otherwise reduced in those tests.

Of course, I'm more strict about such things due to my profession. OTOH this also means I know the consequences of not being careful what it is actually measured vs. what you think it is

All state of the art sensors have, more or less, the same efficiency. So, to improve the noise to signal ratio you have to do noise reduction. It is as simple as that. And those measurements looks very genuine. There are others that do measurements like this. And they generally get the same kind of results. The jump you see in the K-1 II graph is what happens when the camera kicks in noise reduction.

And note, there is not content in the image for this kind of tests. You often just take a set of exposures of a gray card, measuring the clipping and the noise content. From this you can compute the lowest level with a certain signal to noise and also the highest level where it doers not clip. And, from that you can draw those charts.

And that's the problem. You get some nice charts about content-less images.

Alex