[BigMackCam]

Folks, let's keep this friendly and respectful, please. First day of a brand new year, and all that

Thank you!

[biz-engineer]

The assumption here (it's a popular one) is that processing is completely content unaware; perhaps a simple filtering.

Not sure. With Silkypix 8, I've noticed that smoothing amount depending on the level of details of the image. If I have an image with lots of bokeh (blurred background), the noise reduction is stronger in the blurred area that it is on the subject details. When ISO is not ideal, I can still get a better looking overall image by setting an optimal level of smoothing: at the optimal setting, the bokeh becomes smooth (noise free) and the subject still contains most of original detail. So overall , image quality is improved over the original image. If we look at the high ISO test charts from the K1II, we seen that the amount of blur isn't homogeneous, areas with less details are more blurred than areas with sharp details, so I suspect the the accelerator chip is more than just a 2D F.I.R filter, but more like an adaptive 2D digital filter whose coefficient self tune depending on image content.

[Rondec]

RAW purists are not concerned with people not shooting RAW.

Personally, I'd have no problem at all with the "accelerator" processing being mandatory for JPEGs. I'd think it would be unreasonable to not give JPEG shooters a choice as well, but when someone shoots JPEGs they are accepting that the camera will be making irreversible processing choices. So a bit more "accelerator" processing added on top of the regular JPEG processing would break that paradigm at all.

RAW files, on the other hand, are meant to preserve the full development potential. Any destructive manipulation of RAW data compromises this express purpose of shooting RAW.


Surely, we have seen such degrees of detail and cleanliness before. All it takes is to post-process a RAW image.

Personally, I think the hair in the images doesn't look natural unless there is some kind of blur.
Others may say that the unadulterated noise of an ISO 8000 image doesn't look "natural" either.

Either way, in my view it is not possible to state that there is nothing wrong/unusual at all about this image, independently of what ISO level it was shot at.

ISO 8000 shots have less detail. That is true if you try to get rid of the noise in the image. It is true if you don't. Ideally you don't shoot at that iso, but here I was shooting with the 24-70 f2.8 at f3.2. As I say it was low light and I wanted a photo of my daughter with her two cousins. It is not art. There is some detail lost, but it is certainly printable at probably up to 8 by 10 without anyone being bothered by the quality of the image.

That said, I still have not heard the situations where you think that the detail lost would truly affect the image. Astro photography is the only thing that has been mentioned and yet, reh quoted the astrophotography site as saying that they did not see detail loss with the K-1 II images, although it technically could happen. If that's the only situation where this actually is problematic and even dedicated astro photographers can't tell for sure if it is an issue, surely it is pretty small.

Let me say that I have no problem with anyone saying that high iso photography is a balancing act. We try to limit noise while not smudging away too much detail. Some people have a pretty high tolerance of noise (particularly folks who began in the film era). Some have a pretty low tolerance to noise in their images and they tend to be much more aggressive with noise reduction. If you do any noise reduction to your K-1 files in post, you will see some detail loss, regardless of the noise reduction engine you use. I would add further that noisy images often appear to have more detail, but some of that is merely a factor of the noise in the image.

Regardless, to me, reviews of the K-1 II should state clearly that there is an accelerator. It will reduce color noise at the expense of a small amount of detail. This reduced detail is unlikely to be noticed at most viewing sizes. To me, that would be enough to state the situation at hand. There are plenty of people who would opt for a K-1 original (they are still available on the market), but there are plenty of folks who would choose the K-1 II for its perceived benefits.

[reh321]

39-1/2 years ago, I and my bride went to a local bank to open a checking account. For a number of minutes, she {soon to be a PhD} was wasting her time, because the Assist Bank Manager asked me question after question, but ignored her; she was part of the conversation only when I directed a question at her. Why this trip down Memory Lane? If I understand correctly, we in the non-Oriental world play a similar role in the World of Pentax. We are welcome to use their cameras, but our desires are at best tie-breakers in making decisions. Thus, the discussion several pages ago

I keep asking about the preferences of Japanese photographers and keep not getting a clear answer.

Well I, for one, don't know their preferences.

shows why this entire discussion is a waste of time. If it had occurred in June or July, it would have been useful in raising the post count, but now it is just a waste of time. If we don't know what the Japanese market is like, then we cannot understand the decisions Pentax makes; they do not design cameras for DPR staff. The Pentax designers clearly understood the pin-outs of the sensor and the processor, and designed the 'accelerator' to fit in the gap between them; the Pentax designers understood what software could do for their users, and how their Japanese users wished to use this software.The Pentax designers clearly understood their goals and are proud of having reached them. I am immensely enjoying my KP because my goals are in agreement with their goals - I fit into one of the niches they choose to address .... I am sorry that neither DPR nor @Class A understands that niche. I doubt I will ever purchase another camera, but I look forward to what Pentax delivers this New Year just because.

[biz-engineer]

I'm personally considering an email writing capaign to Amazon pointing out that with it's lack fo brick and mortar sellers Pentax is prime candidate for Amazon on-line sales, and that their selling arm does everything it can to discourage people from buying Pentax.

Amazon policy is to provide customers with information such that customers can make the right purchase decision based on what they want to do with the product. It is possible that dp reviews has some bias due to the reviewers being used to use personally some camera models. That may change in the future when the managers who are responsible for taking care of the mission take care to avoid too much bias.

---------- Post added 01-01-19 at 15:32 ----------

That's not correct as far as the RAW data is concerned. The vast majority of cameras scale the values so that the most significant bits are used (provided there is sufficient light)

How do I have move room for boosting exposure in post without blowing highlight, when the RAW was shot at ISO200 vs 100, or ISO400 vs ISO200 etc. My experience with shooting images with sun in the background was that I can still increase exposure in post more when not using base ISO. And that 's exactly how the High light protection feature of the K1 works, it sets min ISO to 200 when the AE meter detect a bright sport in the scene, with has the same effect as underexposing 1 stop at ISO100, except the JPEG preview of the ISO200 shot looks correctly exposed. When shooting at ISO100, the light levels are coded over 14bits; ISO200 => 13bits; ISO400 => 12bits; ISO800 => 11bits ; ISO1600 => 10bits etc. In post, the bits are shifted which restore luminance level appropriate to vision, the extra bits on tone information that was never recorded is never reconstructed by bit shifting and noise reduction (no matter what charts can say about it). Noise reduction doesn't improve dynamic range, but it may improve how smooth the image looks to human eyes.

[normhead]

Amazon policy is to provide customers with information such that customers can make the right purchase decision based on what they want to do with the product. It is possible that dp reviews has some bias due to the reviewers being used to use personally some camera models. That may change in the future when the managers who are responsible for taking care of the mission take care to avoid too much bias.

Ya, we can probably push that change along a little.

[Kunzite]

Folks, let's keep this friendly and respectful, please. First day of a brand new year, and all that

Thank you!

I'm sorry if I was disrespectful in any way. I assure you that I do respect Class A despite strongly disagreeing with him on this specific subject.

[Rondec]

I've posted this before, but this is a night shot of New Orleans at ISO 12K on the K-1 II. Shot at 1/60 second, f2.8 and 36mm with the DFA 24-70mm.



I wouldn't print this image to a huge size, but once again, I think it would tolerate 8 by 10.

[biz-engineer]

ISO 8000 shots have less detail. That is true if you try to get rid of the noise in the image

The whole dilemma about noise reduction is that it is assumed that less noise is better, but it is not necessarily the case. I once showed a B&W print to a friend and he did not like it, he said "Would it be possible to have some grain, because black & white remembers me vintage, and I like images with grain". I can't say noise reduction over detail is better, it depends. I guess that's why Ricoh did not discontinue the K1 when they released the K1 Mk II, so that customers have the choice. IMO, offering more choice to customers isn't bad.

[Breakfastographer]

You are welcome to try replicating the look of a K-1 II image from a K-1 image. You don't know what it's done, don't have the algorithms, likely don't have all the data.

So I went back and tried to do this precisely. Here's what I'm finding.

The chroma splotchiness caused by the accelerator cannot be replicated in PRIME. So I tried it in HQ.

To make the complexity of the settings manageable, I used only the luminance and chrominance sliders and set everything else to zero.

Having done that, these settings get me close:

HQ, luminance 40, chrominance 5

This result is based on matching noise patterns, not detail. The K-1 II image appeared sharper, but even the HQ denoised image retained more detail in places than the untouched K-1 II image - reference image here, my sample below:

[Class A]

That is a very weak argument for rejecting what he's saying and promote uninformed guessing instead.

Why is it a weak argument to observe that someone has a conflict of interest?

Would you expect a Pentax employee to do anything else but let the camera appear in the best light possible?

How am I promoting "uninformed guessing"?
This is a strong allegation and I feel you need to be able to support it.

I repeatedly said that it is likely that there is a translation issue. I never said "this Pentax employee is deliberately trying to mislead us all", or anything of that nature.

I only said that if you take the message verbatim, at face value that it doesn't make sense to an expert.
You don't seem to be an expert in the matter which is fine, but you shouldn't be accusing experts of "promoting uninformed guessing" when they simply point out incorrect statements.

As Breakfastographer just posted about PRIME again, which is based on denoising before demosaicing, I thought about the possibility that the Pentax employee perhaps meant that the image processing they are performing with the "accelerator" unit would not be possible on already demosaiced images. That would be a statement that could be accurate and it is possible that the intended message was lost in translation or just worded in a manner that either intentionally or unintentionally allowed other interpretations.

I think very highly of Pentax engineers, they have my utmost respect because they earned it. I would never just lazily claim that that they are lying just to increase sales.

Having said that, it wouldn't be the first time that Pentax let out technical detail that turned out to be not true. Everyone thought that the K100D already had three axis in-body image stabilisation. It turned out that it only featured two-axis stabilisation (with no support for roll compensation), a fact that was revealed to be caused "by a translation error back in 2006". Again, intentional or not, incorrect technical information was assumed up until 2011 and there is nothing to suggest that this could not have happened again regarding descriptions of the "accelerator" unit that seem to be suggesting more than what the measurements and observations are showing.

you have no desire to support your claim,

I've supported my claims multiple times in many ways. You cannot dictate when a claim has been supported by arbitrarily setting a goal (mimicking K-1 II behaviour) while ignoring objective evidence and expert assessments of the technical foundations.

Sending me to Wikipedia.... that's cute...

That is your response to the question of whether or not I have been correct regarding computability all along or not?
This is now way of having a debate and I don't intend to continue it at that level.

My premise is that I'm looking at the K-1 II images and they look better to me than the K-1 ones, and I don't get why people are so aggressively against it.

I haven't witnessed anyone being aggressive towards you just because you happen to like the K-1 II processing.
People liking the K-1 II are not of concern here.

The real problem is a "Pentax put a lot of work into the processing and on top of that I like it" attitude that demands that everyone ought to accept it as well and is denied the option of opting out of the processing. I've asked multiple times where the problem with giving users the choice to opt out is and have never received a good answer.

But you know what? Let's say, hypothetically, we adopt the notion that the requirements of some Pentax users don't matter and assume it is fine to push the mandatory RAW processing down their throats. It still wouldn't be a smart thing to do because it means handing any review site that intends to demonstrate that they can be tough reviewers the ammunition on a silver plate. Not smart. The potential of losing sales through the bad press the K-1 II got from DPReview could have easily been avoided.

[reh321]

RAW purists are not concerned with people not shooting RAW.

Personally, I'd have no problem at all with the "accelerator" processing being mandatory for JPEGs. I'd think it would be unreasonable to not give JPEG shooters a choice as well, but when someone shoots JPEGs they are accepting that the camera will be making irreversible processing choices. So a bit more "accelerator" processing added on top of the regular JPEG processing would break that paradigm at all.

RAW files, on the other hand, are meant to preserve the full development potential. Any destructive manipulation of RAW data compromises this express purpose of shooting RAW.

You are assuming that "full development potential", whatever that is, is a goal. Maybe the goal is getting the 'best' (*) recording of what I saw {and when recording a meadow scene, I didn't see the ladybug sitting on a leaf even though it might be recorded as a single orange pixel on the sensor}.

The 'accelerator' comes between the sensor and the processor - it doesn't naturally 'know' what kind of file(s) the processor is going to create {in fact, the processor doesn't even know that, since my KP allows in-camera development after-the-fact} - so this idea of turning the 'accelerator' on/off depending on user expectation would be a problem even if it were physically possible.

(*) where NR is included in my definition of 'best'

[biz-engineer]

This result is based on matching noise patterns, not detail. The K-1 II image appeared sharper, but even the HQ denoised image retained more detail in places than the untouched K-1 II image - sample below:

If you take a zone of the image where there is little deterministic details, the built noise reduction may wipe out details as well as noise. But that's because if you look at an image from further away, the image with less noise in backgrounds looks better. There is a reason why subject separation in used in photography. If you want to compare noise reduction algorithms, the image pattern used for comparison should contains plenty of subject details.

[Rondec]

The whole dilemma about noise reduction is that it is assumed that less noise is better, but it is not necessarily the case. I once showed a B&W print to a friend and he did not like it, he said "Would it be possible to have some grain, because black & white remembers me vintage, and I like images with grain". I can't say noise reduction over detail is better, it depends. I guess that's why Ricoh did not discontinue the K1 when they released the K1 Mk II, so that customers have the choice. IMO, offering more choice to customers isn't bad.

True. Although digital noise has a very different appearance from film grain. I like film grain's appearance a lot better than I do digital noise.

Regardless, we are viewing images at 100 percent and comparing noise patterns at those levels. While it is an interesting academic exercise, it does not indicate real world differences.

[Breakfastographer]

You are assuming that "full development potential", whatever that is, is a goal. Maybe the goal is getting the 'best' (*) recording of what I saw {and when recording a meadow scene, I didn't see the ladybug sitting on a leaf even though it might be recorded as a single orange pixel on the sensor}.

Fair point - that's kind of the difference between the Scott Kelby school of thought ("remove in post") and "authentic photography" as I've alluded to here.

The 'accelerator' comes between the sensor and the processor - it doesn't naturally 'know' what kind of file(s) the processor is going to create

What would be some "kinds" of file according to your meaning?

---------- Post added 01-01-19 at 08:04 AM ----------

If you take a zone of the image where there is little deterministic details, the built noise reduction may wipe out details as well as noise. But that's because if you look at an image from further away, the image with less noise in backgrounds looks better. There is a reason why subject separation in used in photography. If you want to compare noise reduction algorithms, the image pattern used for comparison should contains plenty of subject details.

Yeah, did you look at my original post, or just the little NR window excerpt?