[robjmitchell]

Because other manufacturers are doing the same thing these days (see the A7R "star eater" issue). Because the KP uses a known sensor that isn't exclusive to that camera model and there's no explanation for the "better" high ISO in that particular model aside from software baking (which is exactly what the accelerator chip is there to facilitate).

I agree that in actual use, if Ricoh managed to bake in noise reduction that has minimal-to-no impact on image detail... great. Enjoy it. But that isn't my point. My point is that it's disingenuous to compare baked RAW files from camera X to largely unmolested RAW files from camera Y - then go on and on about the amazing high ISO performance of camera X. That high ISO performance wasn't achieved by a superior sensor, shielding, cooling, etc. design. It was achieved with noise reduction algorithms. That's like comparing an untouched RAW to something you've run through Lightroom and reduced noise on. It makes no sense.

That's why I said it's absurd to try to say a KP might "match or best" FF in IQ. That's just absolutely not true. Suppressing high ISO noise to the level of an unmolested FF file doesn't mean it's matched it in image quality. If the sensor was truly better and truly matched FF image quality, we'd see increased DR across the range too. The D7200 was able to achieve an incredible 14.6 stops years ago, so it's definitely possible. At last check the KP and A6500 shared a sensor, and the A6500 is a full stop down from that. Otherwise, I can just take K3 files, run them through Lightroom, and two seconds later I've achieved the same "image quality" as the KP.

Traditional NR is algorithm based, it is destructive as it smears the pixels together to remove the noise. The noise still becomes part of the final picture. Dedicated NR software is just better algorithms. Its essentially an analog solution. The Accellerator chip on the other hand appears to work on a pixel level by altering the individual pixels that are noisy vs the surrounds. It’s essentially a digital solution that stops the loss of dynamic range and detail as it removes the noise before it becomes part of the image.

[normhead]

Traditional NR is algorithm based, it is destructive as it smears the pixels together to remove the noise. The noise still becomes part of the final picture. Dedicated NR software is just better algorithms. Its essentially an analog solution. The Accellerator chip on the other hand appears to work on a pixel level by altering the individual pixels that are noisy vs the surrounds. It’s essentially a digital solution that stops the loss of dynamic range and detail as it removes the noise before it becomes part of the image.

As someone who has a basic understanding of programming, having programmed in machine language int hold days, I have always wondered why this isn't done by people like DxO , photoshop etc. I can think of at least one way of doing it that would totally not affect detail. But I'm not in the software business. All it would take would be taking every pixel that is completely surrounded by very similar pixels and changing the "noise" to the average of the surrounding pixels. I'm sure it would be a lot harder than that, but, there is absolutely no reason for destructive noise reduction, except that folks are really lazy, and other people are willing to buy their product.

[ffking]

Pentax is what it is. People make their decisions, buy what they think is right, then blame Pentax because it isn't what they want.

I think you talk a lot of sense here (the whole post), but there is also the emotional, non-rational aspect amongst some of us - it's like being a long-time supporter of a team - you start to identify with them and get an unrealistic emotional high when they do well and are plunged into despair when they do badly, It's not that Pentax are not doing what they always did ( though they have been quite slow about it of late) - it's just that people (myself included, I have to admit) want to see them back in the premier league where their proud history suggests they should be and sometimes get frustrated. Not rational or hard-headed, just the way some people are - though I do try to stop my personal frustration from spilling over into lashing out and Ricoh or Pentax - or other forum members for that matter - that's not not helpful, as you say.

EDIT - I suppose what I'm saying is that (in sport) the biggest moaners are also some of the biggest cheerleaders when things go right for their team - but yuh - that doesn't excuse choosing to back a low placed team and complaining they don't win all the trophies

[Rondec]

What is Pentax good at?

They are good at making weather sealed cameras with good ergonomics, 100 percent coverage pentaprism optical viewfinders, decent (but not awesome) specifications, and good value. They are pretty good at figuring out ways to leverage their in body stabilization to allow for extra features like pixel shift and astro tracer. They squeeze every ounce of image quality out of sensors that are a little older than the competition and they remain very conservative in their goals going forward.

They aren't good at making spec monster cameras, tracking auto focus, or top end video. They have chosen not to go after the wildlife or sports markets -- probably not because they couldn't do it, but because they think the investment to produce such cameras and lenses would not be worth the return.

Can Pentax make things work going forward?

I think so, but times are going to be tough for a lot of the camera makers out there. Nikon is reshuffling everything. Olympus is doing better now, but they had some dark days not so long ago. But I don't know that Pentax trying to make cameras that don't fit their strengths will work for them. To me it is more about doubling down on their strengths and making cameras that are uniquely Pentax. Leica has done this and done it well. I believe Pentax could and should do it too.

I've been repeating this mantra a lot, but the whole point of camera gear is to create images -- images that you are satisfied with. Whatever else you say about them, Pentax cameras still function pretty well as image making machines and you could do a lot worse than buy a K-P and a set of DA limited lenses. The biggest limitation for most of us is not the camera at all, but lies a couple of inches behind the viewfinder (whether it is optical or electronic is unimportant).

---

Last edited by Rondec; 02-16-2018 at 04:12 AM.

[house]

All it would take would be taking every pixel that is completely surrounded by very similar pixels and changing the "noise" to the average of the surrounding pixels

You are joking right!? What you are describing is the most basic noise reduction there is. Actual in use noise reduction works on that principle but with very complex algos.

The above analog/digital comments are also brutally confused.

[Kunzite]

Isn't that just a median filter noise reduction? Anyway, it's hardly impressive - and definitely not without an impact on detail.

[reh321]

Isn't that just a median filter noise reduction? Anyway, it's hardly impressive - and definitely not without an impact on detail.

Has nobody tested the KP for detail retention in the past year??!!

[RonHendriks1966]

Has nobody tested the KP for detail retention in the past year??!!

With testing sites like dxo being American and earning their money with sales true links........logically they didn't test the KP, since there where hardly any sales in the USA. Sales for Ricoh-Imaging collapsed last year in the America's with a down of 50 % or so.

[reh321]

With testing sites like dxo being American and earning their money with sales true links........logically they didn't test the KP, since there where hardly any sales in the USA. Sales for Ricoh-Imaging collapsed last year in the America's with a down of 50 % or so.

Several years ago I purchased a Plustek scanner; I noticed right away that the 'ICE' feature was erasing writing, not just slide defects. It's too bad that no users here have been watching for disappearing detail with the K-70 and KP.


Added: I stopped using that feature for awhile; a later software upgrade seems to have fixed it.

[Kunzite]

It's too bad that no users here have been watching for disappearing detail with the K-70 and KP.

They were too busy being impressed by the detail retention at high ISO.

[AyeYo]

You are joking right!? What you are describing is the most basic noise reduction there is. Actual in use noise reduction works on that principle but with very complex algos.

The above analog/digital comments are also brutally confused.

That's exactly what I was thinking. That's literally just averaging out the noise... which is smearing it out. That's the polar opposite of detail retention.

You can attempt to retain as many details as possible, but unless you've created software with psychic abilities to know what color and tone that pixel should have been, you're going to be losing detail when doing noise reduction. There's no such thing as "noisy pixels" in the sense you can just clean the noise off that pixel. The information isn't there. They should have been color and tone unknown, but are instead color and tone Y... and there's no way to know for sure what color and tone it should have been.

Truly improved high ISO performance comes from not having that incorrect data in the first place. That doesn't mean averaging it out in software before writing the file. It means getting the correct data from the initial sensor read. That isn't what KP is doing. That isn't what any noise reduction algorithms do. That's what better sensors, better shielding, and better cooling do.

[The Squirrel Mafia]

It's too bad that no users here have been watching for disappearing detail with the K-70 and KP.

From playing around & comparing lots of low ISO to high ISO "studio scene" RAW files from Imaging-Resource, DPR, & other websites in RawTherapee & DxO 11, about the only thing I have noticed with the K-70 in particular is that as the ISO climbs, the color begins to fade out when compared to the K-3II or even the K-50. It does hold on to fine detail better than the K-3II, but the color definitely starts fading out when the ISO gets higher. You can start to notice it at ISO 6400, more noticeable at ISO 12800, & really obvious by ISO 25600. Like the color gets washed out. I'm guessing that the K-70 was used as the test platform for that new chip. If I had a K-70, I'd still set the auto ISO from 100 to 25600 & basically turn images after ISO 12800 into black & white.

On the other hand, the KP is mind blowing up to ISO 25600. My guess is that the engineers kind of figured out the chip even better. It holds on to fine detail & color all while reducing luminance & color noise. Shooting at ISO 25600 on the KP is like shooting ISO 12800 on the K-3II & K-50. It's a stop better. If I had a KP I'd set the auto ISO from 100 to 51200 & turn ISO 25600 images & above into black & white.

I'm guessing that the K-3II replacement will image quality that will be equal to the KP or slightly better.

[Kunzite]

You folks dismissing the KP for "baking" RAWs, did you test the actual camera?

[normhead]

You can attempt to retain as many details as possible, but unless you've created software with psychic abilities to know what color and tone that pixel should have been, you're going to be losing detail when doing noise reduction.

That's why basic noise reduction should concentrate on the areas where there is no detail. Look at Kengoh's technique for his birds in flight. He has to have a high shutter speed and ISO to freeze the birds, so he goes to a spot where they will land, take as few 100 ISO images. Cranks up the ISO and then puts birds in flight onto a clean background. Any noise on the birds is handled separately. I see no reasons why cleaning up the noise in out of focus areas shouldn't be done right off the camera. If there is not strong detail, there is no need to preserve it. The actual subject matter should be treated differently. The fact that you apply most NR algorithms across the whole frame right from the start means you're doing the wrong thing.

---------- Post added 02-16-18 at 11:08 AM ----------

You folks dismissing the KP for "baking" RAWs, did you test the actual camera?

What? You can actually like use a camera before you comment on it's weaknesses? Who knew?
I thought you were just supposed to say the first stupid thing that came into your head.

99% of the time when I see this type of post, like the K-P baking it's files, if I can test it myself, more often than not, I find the percieved fault is meaningless in real world comparisons, where someone is obsessing about something that the average guy wouldn't care about, and sometimes, I find the opposite is true. I actually like the images affected by the supposed weakness more than I like the images without it. Taking these kinds of statements at face value is really risky.

[beholder3]

Has nobody tested the KP for detail retention in the past year??!!

You just need to look at some sample images:
Studio shot comparison: Digital Photography Review
Clear b&w lines are best.
You only need to use your own eyes and check for which camera you see the most clear line separation the farthest to the left, with as little moire as possible.
The KP looses the least detail with the Fuji XT2 being 90% similar. Very closely the K-3II follows and the Nikon clearly lags behind (same with Canon 80D or Sony 6500).

From what you can get out of Lightroom the KP seems the Queen of APSC noise and detail retention.