[Workingdog]

I did not think there would be a significant difference in the way different RAW conversion software would render noise. I was wrong. The 3 crops below all from the same raw file, ISO 2500, K7. The software: 1. Pentax 2. RAWTherapee 3. UFRAW Even more interesting (and subject of a future post) is the way Topaz denoise handles each of the conversions.
Which do you think looks better? Which do you think will be best after Topaz Denoise?
.
.
.

---

Last edited by Workingdog; 12-30-2011 at 06:33 AM.

[Workingdog]

A little more info: In-camera noise reduction OFF. Noise reduction features of all 3 RAW converters set to zero or disabled.

[Laurentiu Cristofor]

Your results here are about the same as those I got when I tried these converters with default settings a couple of years ago. My conclusion was that the PDCU software works very well. But I've also seen some recent conversions using RAW Therapee that looked quite excellent - maybe it's a matter of tweaking the converter settings. PDCU is underrated though.

[hcc]

+1 with previous post. PDCU works quite well, espcailly with the noise reduction sliders from version 4.3 (I believe).

I also use Noiseware and I feel personally that Noiseware give a nicer finishing touch than PDCU, but this is quite personal IMO.

[jac]

NIK's define is great. No reason for me to look elsewhere but I haven't worked withe the others.

[Workingdog]

I must repeat, the 3 conversions have absolutely no noise reduction whatsoever applied. I made sure I turned noise reduction off in all 3 conversion software. My observation is that the Pentax software produces a cleaner image even before any noise reduction.

So, why would the three software render the noise so much differently?

[Class A]

So, why would the three software render the noise so much differently?

It seems that the different converters have different ideas about what to do when NR settings are set to zero.

The Pentax conversion clearly uses quite a bit of chroma noise reduction and some luminosity noise reduction as well.

If you really turned every kind of noise reduction off then that simply means that a "zero setting" does not guarantee that the Pentax converter will apply no noise reduction at all, in particular at such ISO levels.

As an analogy, a number of cameras will still apply NR to RAW data beyond a certain ISO level even when NR is "off".

[Workingdog]

It seems that the different converters have different ideas about what to do when NR settings are set to zero.

The Pentax conversion clearly uses quite a bit of chroma noise reduction and some luminosity noise reduction as well.

If you really turned every kind of noise reduction off then that simply means that a "zero setting" does not guarantee that the Pentax converter will apply no noise reduction at all, in particular at such ISO levels.

As an analogy, a number of cameras will still apply NR to RAW data beyond a certain ISO level even when NR is "off".

I would have to agree with you but never would have thought that it worked this way. Check out the part 2 thread after denoising.

[Laurentiu Cristofor]

So, why would the three software render the noise so much differently?

Because they use different demosaicing algorithms to produce the image. Different algorithms, different results.

[Class A]

Because they use different demosaicing algorithms to produce the image.

Different demosaicing algorithms differ in the sharpness, spurious patterns and colour artifacts they produce. No way changing the demosaicing algorithm alone would explain this difference in noise levels.

[kh1234567890]

Any chance of posting the RAW file somewhere - I'd like to see how Silkypix compares ?

[Laurentiu Cristofor]

Different demosaicing algorithms differ in the sharpness, spurious patterns and colour artifacts they produce. No way changing the demosaicing algorithm alone would explain this difference in noise levels.

Why not? Demosaicing can include some level of NR. You are not surprised about differences in spurious patterns and color artifacts but you find the different rendering of noise to be unexpected?

[wildman]

Any chance of posting the RAW file somewhere - I'd like to see how Silkypix compares ?

Good idea - I must have at least 4 or 5 programs I could try.

[Workingdog]

Good idea - I must have at least 4 or 5 programs I could try.

If both of you gentlemen send me a PM with an e-mail address, I will gladly send you the RAW file to play with. BTW, did you check out the post titled Part 2 to see how Topaz Denoise handles each of these conversions?

[Class A]

Why not? Demosaicing can include some level of NR. You are not surprised about differences in spurious patterns and color artifacts but you find the different rendering of noise to be unexpected?

Demosaicing is the process of reconstructing colour information. The various algorithms all have to use interpolation but differ in what sensels to choose for interpolation, given a certain scenario.

The variations in the algorithms will cause variations in the aspects I listed. They won't impact significantly on noise. None of the papers I've read about demosaicing algorithms (AHD, etc.) mentioned integrated noise reduction. EDIT: I just searched and found a research paper that proposes to combine demosaicing and denoising. Not sure if this has been taken up yet, but in any event, the denoising part should be adjustable, i.e., it should be possible to reduce denoising to zero, should you wish to do so.

Note that I'm not talking about "...different rendering[s] of noise...". The sample images clearly show different levels of noise, not different renderings of noise.