Noise pattern with Fuji X-trans RAW files

(Sebastien Guyader) #21

I’m sorry for the language, my high-school Deutsch is now limited to just a few words. But I think we under understand each other well enough in English.
I cannot say the settings I used ar the perfect settings for iso 6400. You have to try and see if it satisfies you.
Maybe you can lower noise reduction and micro-contrast with ISO 1600 images, but here again you have to try.


(Michael) #22

With your help, I see good progress in developping with RT. Again: thank you very much.

Without noisreductin, the image has so ugly “pixel-blocks” or artefaktes: 3, 4, 5 or 6 pixels together in horizontal or vertikal direction. This looks not organicaly. What is the reason? Is this a RT or a Fuji-X-trans probelm? I have seen, with Lightroom it is simmilar. Is this some kind of mistake or a normal phenomen of digital fotografie? Why are this smal “pixel-blocks” not seen when I use fast-demosaic in RT?

Does somebody know, whether Lightroom is fundamental better than RT? (I dont like to use Adobe, I am satisfied with gimp and co.) Can I have about the them results whit RT? That would make me very glad.


(Alberto) #23

it all depends on the way you measure quality. without a clear definition, we can argue until the end of time. it is also very likely that the answer is highly subjective.


Unfortunatly i don’t understand german…

For what can i say, with xtrans rawtherapee is what i think the best converter (and not only with xtrans) :wink: .

Only advantage with lightroom with xtrans is better/much easier distorsion-vignetting-chromatic aberration correction (but as soon as rawtherapee will get lensfun integration, rawtherapee will win in this area too).

(Mica) #25

Lightroom and RT are just tools, and the best tool is the one you can wield skillfully.

(Guillaume Castagnino) #26

Yes, but beware, it drastically reduces the colour resolution. And the more the image is noisy, the more it destroys colours.
So use it only when really needed!

I agree 100%. This is why I always shoot raw with my X20 (very small X-Trans sensor) and X100s. I was constantly disappointed with the noise reduction level, even set at minimum.
A little luminance noise is better than too much noise reduction. This allow to keep better local contrast and details that are usually flattened by NR algorithms

(Michael) #27

The quality, I am looking for is, how much the ugly artefactes are vilsible, as shonw in the jpg-examples bevore.
When I trye to eliminate this “pixel-blocks” I have to use NR, and NR makes the image unsharp and smeary.
This was my question: Is RT similar to LR, or is LR more capabel to produce a organiy looking picture?

(Alberto) #28

I have no experience with xtrans nor lightroom. but the wise of internet frequently complain about the poor support for xtrans in Adobe products, mentioning RT as a good alternative. so I suppose there’s no fundamental advantage of lightroom here. although I 100% agree with @paperdigits: the best tool is the one you can master better…

(Sebastien Guyader) #29

@micha I also think you can lower the noise level by carefully setting your shooting conditions. ISO 6400 like in the example you provided is always to look very noisy. Also in this example phot, I had to dial in +1.6 EV, which makes the noise even more prominent. So, try to avoid shooting at such high ISO and underexposing.

(Sebastien Guyader) #30

I agree. I generally avoid using it, and I prefer having some artifacts (barely visible unless pixel peeping) than color smearing.


I find that quite a few demosaicing methods, e.g. AMaZE, have a crosshatching pattern. I have grown to accept it but it can be annoying when shooting conditions aren’t ideal and you end up with an extremely noisy image. Sampled from a random noisy raw (not RAF):




The crosshatching above is fairly mild, likely because of my unsteady hands; often the hatches are sharper and more distinct. This can be disturbing for Lightroom/ACR users but, when you stop pixel-peeping, FOSS solutions win in terms of resolution.

(Michael) #32

Hallo Sebastien,
I have now learned, (from you and the others in this blog), how to handle with noisy raf. In my examples I have chosen the high ISO, to have a bad “negativ”. Normaly I try to avoid this, but sometimes, I like to take fotos with littel ligth.

My most importent question is: What is the reason of this artefactes “pixel-blocks” (5 to 12 pixels in a ligne), you can se in my examples and also in the picture of the books from heckflosse. Is this a normal phenomen by digital cameras? Or is this a Fuji-problem. Or is this a RT-problem? Or, is this no problem, because in a normal foto-print, you do not se this?

(Sebastien Guyader) #33

Hallo Michael,
I don’t have a broad knowledge in demosaicing. We would need to compare similar high shots from xtrans and Bayer sensors, and process them with similar settings in RT to see if it’s an xtrans problem. I very confident that it’s not a bad behavior of RT.
Actually you search which cameras share the same Sony sensor but differ with their CFA, and download high ISO shots from sites like or for comparison. Unfortunately these days I don’t have time to do this research.

(Ingo Weyrich) #34

For bayer sensors there are several demosaicers. Amaze is to get most details out of low /medium ISO bayer raw files. For high iso bayer raw files there are lmmse and igv (at least in rt) which result in less artifacts.

For xtrans sensors there’s only 3-pass (gives most detail), 1-pass (less detail, but faster than 3-pass) and fast.
fast should only be used for less than 100% view. 3-pass gives most detail on low ISO shots while 1-pass leads to almost the same output as 3-pass on high ISO shots but is faster.

There is no special demosaicer for high ISO xtrans shots (like lmmse or igv for bayer)

(Michael) #35

Hallo Ingo,
ich hoffe, ich darf in diesem Forum auf Deutsch schreiben. Meine Fragen An Dich sind:
Gibt es die Möglichkeit in RT zwei unterschiedliche Bilder gleichzeitig anzuzeigen? Ich möchte natürlich das in Bearbeitung befindliche Raw-Bild sehen und daneben das Original von der Kamera entwickelte Jpeg (oder ein anderes, bereits entwickeltes Raw-Bild, um die Farben anzugleichen).

Was ist der Unterschied zwischen Belichtungskorrektur und Helligkeit bei Belichtung?

Was ist der Unterschied zwischen Belichtung/Sättigung und Dynamik/Pastelltöne?

Was kann Wavelet besser, wann lohnt es sich, dieses Werkzeug einzusetzen?

Wie schlimm ist es, dass RT noch keine Objektivkorrekturen lesen kann, sondern nur die Verzeichnung korrigiet?

Besten Dank

(Ingo Weyrich) #36

Klar, aber mach nicht zu viele auf, da RT sonst abstürzt. Ab 5.2 wird die Anzahl der offenen Editor-Tabs unter Windows limitiert sein, um diese Abstürze zu vermeiden.

Bzgl. der anderen Fragen am besten mal hier nachschauen. Ist allerdings gerade nicht erreichbar :frowning:

(Michael) #37

Hallo Ingo,
auch wenn ich Mulit-Reitermodus einschalte, ich kann immer nur ein Bild sichtbar machen, und nicht, wie erhofft, zwei gleichzeitig (zum Vergleichen). Was muss ich da anders machen?

(Ingo Weyrich) #38

Sorry, da habe ich dich falsch verstanden. Zwei Bilder gleichzeitig sichtbar machen ist leider nicht möglich.