[tax]

Hello, There are four denoise filters in Darktable (Linux analog of Lightroom) - (1) non-local means, (2) bilateral, (3) profiled and (4) raw. I assume, that the same or similar denoise filters are in Lightroom. Is it possible to achieve the same results that are produced by the accelerator unit in K-1 II at higher ISO's using just mentioned denoise filters on pictures produced with a plain K-1 model?

---

Last edited by tax; 08-30-2018 at 03:21 PM.

[CarlJF]

Denoising in LR is okay but not great. It does fine to remove a low level of noise, but it's better to use something else for heavy noise. However I can't say if the results with a K-1 would be as good as those from a K-1ii. But the accelerator unit had the advantage of giving you the result OOC, without having to PP the RAW.

[Not a Number]

Ricoh claims the effects of the accelerator unit cannot be done with software. Since the processing occurs between the sensor and the image processor there may be a grain of truth in this.

[superdave]

I always prefer denoising in post processing, as it a matter of taste, and if tomorrow there is a new algorithm that outperform the current state of the art, we can still use it in our old RAW files. If a RAW file is denoised, details are lost forever. Moreover software denoiser can ne more efficient as they can tale whatever time they need to cleanup the file, there is no speed vs quality tradeoff.

[tax]

if tomorrow there is a new algorithm that outperform the current state of the art, we can still use it in our old RAW files. If a RAW file is denoised, details are lost forever. Moreover software denoiser can be more efficient as they can take whatever time they need to cleanup the file, there is no speed vs quality tradeoff.

These are very valid points against the accelerator unit.

[MJKoski]

There are already machine learning demos which far surpass everything so far when it comes to denoising images.

One here:

You Tube

You Tube

Another: Learning-to-See-in-the-Dark/README.md at master · cchen156/Learning-to-See-in-the-Dark · GitHub

So, now all is needed is extra computing power in-camera and some material for AI to learn.

[BigMackCam]

Hello, There are four denoise filters in Darktable (Linux analog of Lightroom) - (1) non-local means, (2) bilateral, (3) profiled and (4) raw. I assume, that the same of similar denoise filters are in Lightroom. Is it possible to achieve the same results that are produced by the accelerator unit in K-1 II at higher ISO's using just mentioned denoise filters on pictures produced with a plain K-1 model?

I'm pretty confident you could achieve at least the same quality of noise reduction as (and possibly better than) the K-1II using a K-1 and Darktable.

In addition to the denoise modules you mentioned, there's another - "Equalizer" - which is, in my view, the most versatile and powerful way of addressing both colour and luminance noise. However, it has limits so far as colour noise reduction is concerned, so I tend to use it in conjunction with the bilateral denoise.

For images with relatively low levels of colour and luminance noise, I have one instance of Equalizer dealing with colour noise (blended uniformly for colour) and a second instance dealing with luminance noise (blended uniformly for luminance). For high ISO images, I add one instance of bilateral denoise set to maximum radius of 30 and relatively low levels of noise reduction (generally between 0.050 and 0.150) for each of the channels, blended uniformly for colour.

The benefit of the K-1II is that you're getting noise reduction tailored to each ISO level straight out of the camera. With the K-1, you'd need to apply different levels of noise reduction in Darktable depending on the ISO setting used.

---

Last edited by BigMackCam; 08-30-2018 at 09:24 AM.

[Rondec]

To me the k-1 II is as much about efficiency in editing as anything else. Most shots below iso 6400 on K-1 II do not need any noise reduction.

[rawr]

It has to be said that the biggest advantage of de-noising in post-processing is that it can be done selectively - ie you can choose to only 'paint' NR onto areas of the image that need it in Lightroom, Capture One, DxO PhotoLab etc.

In-camera NR, whether it be the K-1 II or any other, is unfortunately global. The strength of some of the camera NR algorithms can be influenced by stuff like the cameras ISO setting, but they aren't intelligent enough to be very scene selective.

[biz-engineer]

My method to reduce noise is to downsize images. Noise reduction attenuates random signal beyond a certain spacial cuttoff frequency. So in reality, for example for printing, noise reduction by software doesnt bring any additional IQ. Reducing print size or looking at a print further aways acts as the best noise filter you can ever use because the perceived noise reduction match exactly the level of detail that can be perceived.

---------- Post added 30-08-18 at 22:28 ----------

Noise reduction by third party softqare or in camera is good for pixel peepers...

[BigMackCam]

My method to reduce noise is to downsize images. Noise reduction attenuates random signal beyond a certain spacial cuttoff frequency. So in reality, for example for printing, noise reduction by software doesnt bring any additional IQ. Reducing print size or looking at a print further aways acts as the best noise filter you can ever use because the perceived noise reduction match exactly the level of detail that can be perceived.

Are you talking about luminance noise reduction, as opposed to colour noise reduction?

I find that colour noise - especially on higher ISO shots - can be very noticeable even on significantly-resized photos, and leads to inaccurate perception of colour. As a result, if I can see colour noise at 100% reproduction, I'll typically remove it.

For luminance noise, I tend to resize to the required dimensions then decide if any luminance noise reduction might benefit the final image. In some cases - particularly at very high ISO - a little luminance noise reduction can be helpful, even if it's at the cost of a tiny amount of detail. But I'm always very gentle with it. I'd rather have a grainy-but-well-detailed shot than one that's too smooth...

[tax]

I'm pretty confident you could achieve at least the same quality of noise reduction as (and possibly better than) the K-1II using a K-1 and Darktable.

In addition to the denoise modules you mentioned, there's another - "Equalizer" - which is, in my view, the most versatile and powerful way of addressing both colour and luminance noise. However, it has limits so far as colour noise reduction is concerned, so I tend to use it in conjunction with the bilateral denoise.

For images with relatively low levels of colour and luminance noise, I have one instance of Equalizer dealing with colour noise (blended uniformly for colour) and a second instance dealing with luminance noise (blended uniformly for luminance). For high ISO images, I add one instance of bilateral denoise set to maximum radius of 30 and relatively low levels of noise reduction (generally between 0.050 and 0.150) for each of the channels, blended uniformly for colour.

The benefit of the K-1II is that you're getting noise reduction tailored to each ISO level straight out of the camera. With the K-1, you'd need to apply different levels of noise reduction in Darktable depending on the ISO setting used.

Thank you so much BigMackCam!
This is the answer I was looking for. If you can emulate or better K-1 II accelerator unit using Darkroom denoise and equalizer modules, then the only practical advantages K-1 II has over plain K-1 are K-1 II's improved auto focusing algorithm and hand-held pixel shift mode.
I always wanted to have a full control over RAW-DNG images produced by my camera. K-1 in this respect is like Linux - you have a full control over your system but you have to know what you are doing, and "there is no spoon." K-1 II is more like Windows or Apple - there is something going on under its hood, but you don't know, not allowed to know and have no control over it.

[BigMackCam]

Thank you so much BigMackCam!

You're very welcome

This is the answer I was looking for. If you can emulate or better K-1 II accelerator unit using Darkroom denoise and equalizer modules, then the only practical advantages K-1 II has over plain K-1 are K-1 II's improved auto focusing algorithm and hand-held pixel shift mode.
I always wanted to have a full control over RAW-DNG images produced by my camera. K-1 in this respect is like Linux - you have a full control over your system but you have to know what you are doing, and "there is no spoon." K-1 II is more like Windows or Apple - there is something going on under its hood, but you don't know, not allowed to know and have no control over it.

In fairness, the K-1II undoubtedly provides better noise control straight out of the camera. Some people like that, some people don't. For those that like it, then as @Rondec stated, it makes for a more efficient (and faster) workflow. Whether that matters to you or not is an individual thing, depending on your personal preferences and also your application(s) for the camera. I'd guess a wedding shooter would be delighted to have the K-1II over the K-1, for instance.

Personally, I'd be very happy with either the K-1 or K-1II. Given the amount of high ISO shots I take, I may even be a little happier with the K-1II - but then, since I don't take a huge number of shots in one go, maybe I'd be just as happy with the K-1... I honestly couldn't say for sure, except that I'd be happy with either one.

[Pikselisiirto]

I wonder how close K-1 could get compared to K-1 II with software noise reduction. K-1 II should at least have an option to disable them and have parameter adjustments. Ricoh also said that dynamic pixel shift use shake reduction mechanism for additional data, but I haven't seen any other benefits than convenience compared to normal photo stacking.

[BigMackCam]

I wonder how close K-1 could get compared to K-1 II with software noise reduction.

You'll get better results from a K-1 and software noise reduction than a K-1II without (I can confirm this, having worked with test files from both cameras). But, the K-1II provides very good results straight out of the camera. Plus, because the in-camera NR isn't heavy-handed, those files respond extremely well to further software noise reduction too, such that you can achieve at least as good results as (and in some cases, arguably, better) than the K-1.

K-1 II should at least have an option to disable them and have parameter adjustments. Ricoh also said that dynamic pixel shift use shake reduction mechanism for additional data, but I haven't seen any other benefits than convenience compared to normal photo stacking.

Yes, it's been discussed at great length, and I'd have to agree - the in-camera NR should have been switchable. But in real day-to-day use, I'd say most people will see only a benefit from it, and few if any negative side effects...