[reh321]

But fuji doesn't bake nr into raws, as far as I know, .

The term “bake” is used in a pejorative way, but all it means is “cannot be reversed”, but that is true of any change made to the pixels before they are delivered.
Who would want to add noise anyway?
No one has clearly shown that the noise reduction is harmful to an image.

added: we are in the bad habit of judging cameras by specs and other inputs.
We should evaluate them by output, by the images the images they produce.

---

Last edited by reh321; 05-06-2021 at 05:30 AM.

[slartibartfast01]

From examples its clear that Pentax follows the common preference of removing chroma noise whilst preserving luminance noise. I imagine this might be what they are referring to.



Other manufacturers such as fuji are quite aggressive with luminance nr as well. But fuji doesn't bake nr into raws, as far as I know, so any nr is completely optional.



Pentax accelerator nr seems very well judged but that is besides the main point.

The acid test is probably stacking astrophotography images. Hopefully examples will appear soon.

[MMVIII]

The acid test is probably stacking astrophotography images. Hopefully examples will appear soon.

exactly. Any discussion is moot and stays theoretical as long as noone has tested it or used in a certain way and maybe noticed a real discernible disadvantage. I see the theoretical point for special applications as stacking for astrophotography, but by now I haven't heard of anyone having been limited in this field by the nature of these images produced by Prime IV with AU or Prime V with AUII.
And I do think that the image processing clearly has to be seen as one package that is done by this combination. One indication would be its name: accelerator unit, and not NR chip or whatever.

[Class A]

The Pentax Engineers said so at one of the CP+ interviews...

1. The interview you are referring to was held at CP+ 2019.
2. It is not the case that multiple "engineers" said something about the "accelerator unit".
3. A single person, Tetsuya Iwasaki, spoke about the "accelerator unit".
4. Iwasaki-san was not an "engineer" but a manager (Product Planning Department). He may have technical expertise as well but at the time he made the statement about the "accelerator unit", he was a manager.
5. Iwasaki-san expressed uncertainty about the ability to turn off the "accelerator unit". He said (emphasis is mine)

   "However, this issue with the Accelerator is not just a matter of an on/off switch and I think it is difficult to deal with it."

6. It is unclear which difficulties he had in mind. Perhaps he was just considering other functions of the "accelerator unit" (e.g., AF support) and expressed that these other functions would be lost if the whole "accelerator unit" would be turned off. I don't know what he really meant but surely you don't know what he meant either.
7. The statements of a manager are typically not 100% reliable as far as technical details are concerned.
8. Iwasaki-san said

   "But it's also the case that on certain subjects, results without the accelerator unit can be better."

What is so difficult in acknowledging that Pentax themselves recognise that the "accelerator unit" does not always improve images? See the last bullet point in the list above and the quote I posted earlier in this thread.

Pentax themselves admit that the "accelerator unit" can have harmful effects but I constantly have to deal with

1. complete denial by some posters here about the "accelerator unit"'s shortcomings, and
2. accusations I'd be making things up, and
3. claims that "Pentax engineers" said something they never said.

Furthermore, it does not matter whether or not Pentax's programming/design of the "accelerator unit" allows it to be bypassed or not. My point is that a chip like that can be programmed/designed to not change the data flowing through it, if so desired.

If Pentax's implementation does not allow a bypass mode then that's because they decided they don't want one. It is absolutely not the case that such a mode would be somehow technically impossible.

[Class A]

You should also ask those assuming it only does NR, how could they know that

Who says it "only does NR"?

If you are referring to me, I wrote earlier that the "accelerator unit" also affects saturation and sharpens images.

I am not making claims as to whether or not the "accelerator unit" also supports AF or has any other functions. Such additional functions are irrelevant for its image processing function. Image processing could be 100% neutral while other "accelerator unit" functions could still operate normally. It is not a matter of physically switching the whole chip off.

My source is the Pentax engineers as mentioned by reh321.

I see. This means you have no source because not even a single Pentax engineer said something that supports your claim.

[Kobie]

I ran a test last night night between JPG and DNG files with NR at default levels.
I've uploaded them to a folder in my google drive for anyone who's interested. I didn't have time to put the ISO level for each pair but you can get that from the exif anyways.
The raws retain much more detail as ISO increases, and at about 25,600 ISO the jpegs lose color info (no more red on the home theater power button. It's pretty evident that the Accelerator does more than NR, at the least, it does color correction, sharpening and seems better at exposure control of signal output for raw files.
The raw files in the folder are untouched and the jpegs are at default directly from the camera.
Here's the link

[AgentL]

The first thing I do with any digital camera is turn the noise reduction OFF or as far down as it goes (Panasonic "-5"). This has NOTHING to do with the accelerator, in most cameras in fact it only affects JPEGs. But it's still the right idea. Also, Pentax with their RAW pipeline is doing something better than what most of us can do with our editing software, and I wish people would stop complaining about it. My KP and GRIII RAW results are fantastic, I don't care if there's NR being done in a proprietary way higher up in the processing pipeline, because the images look GREAT.

[Kunzite]

[*]Iwasaki-san expressed uncertainty about the ability to turn off the "accelerator unit". He said (emphasis is mine)

"However, this issue with the Accelerator is not just a matter of an on/off switch and I think it is difficult to deal with it."

If not even such a strong negation can convince you...

But, wow, you're even attacking Mr. Tetsuya Iwasaki's credentials, he's just "a manager" so he's "not 100% reliable" (compared to the average forum user I'd guess)
Suddenly he becomes knowledgeable and reliable when he says something convenient!

I still have to see the detrimental effect on actual images. I'm not saying there can't be any... I just want to see examples.
So far the only detrimental effect is some people's extreme insistence on this subject. I'm convinced such people wouldn't be stopped by a mere Off setting... they'd want to make sure every image comparison is done with the Accelerator Off (because otherwise it's "cheating", they'd complain that Off is only partially Off, if someone would present excellent looking images they'd ask, "but was the Accelerator Off, or On?".

[Kunzite]

I ran a test last night night between JPG and DNG files with NR at default levels.
I've uploaded them to a folder in my google drive for anyone who's interested. I didn't have time to put the ISO level for each pair but you can get that from the exif anyways.
The raws retain much more detail as ISO increases, and at about 25,600 ISO the jpegs lose color info (no more red on the home theater power button. It's pretty evident that the Accelerator does more than NR, at the least, it does color correction, sharpening and seems better at exposure control of signal output for raw files.
The raw files in the folder are untouched and the jpegs are at default directly from the camera.
Here's the link

Please do the following test: choose a high enough ISO (25600, why not?), and shoot in RAW two otherwise identical frames, but one with High ISO NR to maximum and another with it disabled.
Any RAW processing would have to be done without any additional NR.

If they're about the same (as I suspect), clearly the option has no influence on the Accelerator.

[Class A]

??6763215 | ???????IP Force?

What parts of the patent description do you think support your views?

I read phrases such as "image data from an image sensor is preprocessed by a pre-processing unit" (the latter is the "accelerater unit"). We know the sensor outputs digital data so the description confirms that there is no online analogue signal processing.

I also note that

"The front-end processor 18 further includes a communication interface 56 that receives control signals from the back-end processor 20. The control signal received from the back-end processor 20 is stored in the register 58, and the front-end processor 18 makes various settings based on the control signal of the register 58."

In other words, as expected, the "accelerator unit" ("front-end processor 18") is configurable, meaning that how it processes data depends on settings set by the DSP (unit 20). Your claim about the impossibility to let the "accelerator unit" behave neutrally, equates to the claim that the DSP in the K-1 II can configure the "accelerator unit" to be neutral for ISO settings below ISO 640 but cannot do so for ISO settings above ISO 640. This is an untenable claim.

[Class A]

That is incorrect.

You very well know of Bill Claff's analysis of the K-1 II.

Denying the fact that the K-1 II performs smoothing of image details is just silly. First, it is visible, even the respective pentaxforums article acknowledged it. Second, the 2D Fourier analysis of the K-1 II images shows at what ISO settings details are smoothed (as a side effect of denoising).

The above has nothing to do with Pentax's own warnings about excessive denoising. There is independent evidence regardless what Pentax say on the matter. Pentax's statements, BTW, support what everyone with a modicum of signal processing knowledge knows; that denoising affects the integrity of photographic data. Your claimed "noise reduction" device that only removes noise but leaves all "signal" intact does not exist. Pentax don't make the claim it exists, why do you?

[Kunzite]

What parts of the patent description do you think support your views?

It's entirety. Or, rather, I'd adjust my views according to information when it becomes available - rather than obsessively insisting on a wrong view.
In this specific context, it shows other processing done by the Accelerator.

Your claim about the impossibility to let the "accelerator unit" behave neutrally, equates to the claim that the DSP in the K-1 II can configure the "accelerator unit" to be neutral for ISO settings below ISO 640 but cannot do so for ISO settings above ISO 640. This is an untenable claim.

That is not my claim; that is a strawman. I never claimed that it is absolutely impossible to disable a certain processing step.
Is it that easy, though? And is it desirable?

[Class A]

The term “bake” is used in a pejorative way, but all it means is “cannot be reversed”,...

What do you mean "all it means"?

That's the problem, that the processing is destructive, i.e., non-reversible.
There is nothing "pejorative" about calling a spade a spade.

...but that is true of any change made to the pixels before they are delivered.

Of course that is not true.

Adding a constant, multiplying by an integer factor, lossless compression, etc. are operations that can be reversed.

Who would want to add noise anyway?

No one.

The point is to not meddle with the capture noise because it contains signal below the noise threshold.

No one has clearly shown that the noise reduction is harmful to an image.

That does not need to be shown, it is known.

If you do not trust your own eyes, it would help to read about signal processing. Consider a crossover component in a two-way loudspeaker. The woofer receives a signal where high-frequencies have been attenuated (the order of attenuation depends on the filter design, a first order design will dampen high frequencies with 6dB/octave). Do you think you can somehow reconstruct the full audio signal (with all its frequencies) at the woofer output of a loudspeaker crossover? Well, you cannot. Once the higher frequencies have been attenuated, they cannot be reconstructed anymore.

Denoising is essentially just applying a low pass filter to image data (rather than audio data).

[Kunzite]

You very well know of Bill Claff's analysis of the K-1 II.

Denying the fact that the K-1 II performs smoothing of image details is just silly. First, it is visible, even the respective pentaxforums article acknowledged it. Second, the 2D Fourier analysis of the K-1 II images shows at what ISO settings details are smoothed (as a side effect of denoising).

The above has nothing to do with Pentax's own warnings about excessive denoising. There is independent evidence regardless what Pentax say on the matter. Pentax's statements, BTW, support what everyone with a modicum of signal processing knowledge knows; that denoising affects the integrity of photographic data. Your claimed "noise reduction" device that only removes noise but leaves all "signal" intact does not exist. Pentax don't make the claim it exists, why do you?

Yes, I know that random guy on the Internet's analysis.
You're wrong, it doesn't show any "details smoothed". It doesn't measure detail, it cannot, not with those test charts. Of course, Bill would make claims about "strong NR" and "smoothing", but I'd say he's overreaching.
His analysis says nothing about any detrimental effect in our images.

Pentax makes various claims, of which the anti-Accelerator people apparently have no knowledge. It's quite telling they cannot even describe which kind of "noise reduction" they want disabled.

[house]

Is it that easy, though? And is it desirable?

It's clearly Pentax judgement that the noise reduction is desirable and that it's not desirable to disable the processing.

I think Pentax vision is that the camera process will be good enough allowing you to spend less time in front of the computer. Fuji is gaining followers partly because sooc is good enough and I think Pentax is going for a similar thing. It's unfortunate that this means more black box, more proprietary and less control for those capable or interested in doing this.