Noise, shadows, etc.

[unfocused]

I'm bored and frustrated by the never ending discussions of shadow noise, banding, etc. etc.

Many of you on this forum talk about how proper post processing and noise reduction can address these problems that others allege are so horrible with Canon sensors.

Let's NOT get into that debate on this thread!!!!! (I fully intend to report and request that the moderators delete any posts that seek to hijack this thread and use it as a soapbox)

Instead, I'd like to know what techniques people use to reduce noise, banding, etc. etc. What post processing tricks do you employ in Raw, Lightroom or Photoshop to fix noise issues?

 

[privatebydesign]

I shoot massive DR subjects on an almost daily basis and it is my living.

I have no time for all the hullabaloo here on this subject but I will offer my techniques.

I expose to the right, generally drop the exposure in post and very occasionally I move the noise slider off zero.

If I bracket I will blend in PS to 32bit files and then use those in ACR or Lightroom. If I get strong blue colour casts in the shadows I might open as a layer the best image for the area, desaturate it or open as a smart object and adjust temp and tint and then mask the shadows in.

My most complicated issues with huge DR are colour shifts from the different light sources which most HDR shooters entirely ignore because everything they shoot is a crazy orange due to them wanting to see detail on the surface of the sun, I deal with these colour differences in a variety of ways that have become ever more convoluted, but I have gotten much quicker at them. For me and the massive DR images I shoot colour shifts are a much bigger problem than low iso noise and banding.

 

[dtaylor]

Instead, I'd like to know what techniques people use to reduce noise, banding, etc. etc. What post processing tricks do you employ in Raw, Lightroom or Photoshop to fix noise issues?

One thing I do is try to minimize noise in ACR before the file is converted for PS. 3rd party plugins have some of their own advantages, but if you can clean it up in RAW it seems to be better in my experience.

So I will tailor the luminance and color NR related sliders to the image while viewing at 100%. If you need to, you can use the adjustment brush to tailor NR in specific areas. Or produce two files from the RAW and blend.

Luminance NR can obviously destroy detail, so you have to be careful, but color NR seems pretty effective and relatively non-destructive.

 

[Keith_Reeder]

Instead, I'd like to know what techniques people use to reduce noise, banding, etc. etc. What post processing tricks do you employ in Raw, Lightroom or Photoshop to fix noise issues?

Well, not being joined at the hip to a single Raw converter is a bloody good start - it shouldn't be news to anyone (and yet it surely seems to be) that some converters are better than others at heading off noise/banding problems, so being open-minded here is a very smart thing: having a few converters under your belt is still a cheaper option than "needing" to go from crop to FF for better noise performance; or "needing" to jump ship to another camera brand so that you can open up the shadows a bit...

I don't actually rate Lr/ACR as highly as I used to in terms of noise handling, and give the nod to Capture One these days. This is what 10,000 ISO looks like (from my 70D) straight out of Capture One at default NR, with no additional PP NR added: and this is a 100% crop of a 6400 ISO file from the same camera, again with no additional NR (and check the Exif - this was not a well-lit scene. "Available Dark", really).

As to "that other thing" (that you've forbidden us to discuss!): this is a 100% crop of a low (160) ISO 70D file off the web somewhere, and this is the same thing out of Capture One, with the Shadows slider maxed out and +1 EC in Capture One. More than 5 stops' adjustment, all told, on the shadows.

At the image level, it looks absolutely fine (starting from this), and (because, you know, everybody's printing big ) it would print really well. All the detail that was in the shadows is there; and nothing's chewed up with tons of chroma and/or banding. That slight texture (visible at 100% view) in what were the darkest shadows is nothing much to worry about.

This is partly down to the camera itself (the sensor in the 70D is actually pretty damn' good - no pattern noise/horrendous chroma in the shadows here); and partly down to the way Capture One converts the files.

(Just for larks, and to emphasise the point that choice of converter matters: this is the same file out of DPP, with only default NR; with the Shadows slider banged over to the right as far as it will go; and with the maximum +2 EC added. Clean as a whistle in the shadows, at zero cost...)

In my own shooting, I used to ETTR with my 7D, which kept things clean (it didn't take much of an adjustment, and I had no qualms at all about using the 7D well into 4-figure ISOs - 3200 ISO here. This is dull, overcast, end of a Winter's day light incidentally, and I ended up at 5000 ISO, IIRC, to maintain a reasonable shutter speed. No problem.

Again, smart choice of converter would do most of the noise "heavy lifting" (Capture One here too, as I recall); but I also habitually used selective sharpening (and where necessary, selective NR) in order to maximise image quality: nothing more complex than a duplicate layer in PS, with sharpening (or NR) applied globally and then erased from where it wasn't wanted.

I don't use ETTR as much with the 70D - it's generally slap-bang in the middle unless I'm shooting in low light, in which case I add some + EC and bump the ISO to maintain shutter speed.

Selective sharpening/NR is still part of my bag of tricks where necessary.

I am a big fan of Topaz Denoise, but to be honest, the native NR in current versions of Photoshop is really good, and Denoise is getting less use than it used to. I do a lot of my selective sharpening with Topaz Adjust these days - I prefer the result to that provided by Smart Sharpen.

It's all very easy...

 

[Sporgon]

The images of mine that are shown on the Building Panoramics site are all, without exception, a maximum of two exposures. I never have trouble with shadow detail and I never have trouble with noise.

There just isn't the EV range in a normal scene that people think; in fact when dealing with incident light, that is the light falling on the subject, the EV range is never going to be more than six to eight stops, and this is within the dynamic range of transparency film, never mind the latest digital sensors.

This all changes when you include the light source itself, and this is where the fighting starts. I don't know what the EV value of the sun's surface on a clear day, but I can tell you it is way, way beyond the D810's sensor !

In my pictures I want to show the sun as I see it, ie I can't see detail in it, or even look directly at it unless it is just before sun set or sun rise, and then I might expose to show detail. I would not consider my pictures to be 'photoshop'd' much at all really, only in so much as to correctly expose for the latitude in the incident lit scene and for the light source - the sky. If you are shooting with the sun or light source behind you then one exposure easily copes with the dynamic range because that range is so within the DR capability of the sensor.

When dealing with scenes of high contrast it's important to start with a raw image that is as flat as possible, that is do not apply jpeg settings to the raw conversion where you might produce a 16 bit TIF file that already has too much contrast, and you then start having to alter the RGB.

The best advice I could give to someone who is wanting to learn about exposure is buy an incident light meter. People can then find out for themselves how the EV range can be positioned on the sensor, and the compromises that then have to be made when you include the light source itself in your exposure. It will also show what the histogram should look like for "correct" exposure, and then how much of a movement you are making in maximising exposure over the range of the sensor. Just remember that the ISO on your camera isn't always what it says it is, so for instance if you have a 5DII set at 100 ISO this is really 73, so I set the meter to ISO 64. 100 ISO on a 5DIII is 80, so I set the meter to 80. Otherwise you'll be thinking 'this meter that that git Sporgon recommended is underexposing' !

 

[scyrene]

I very much tailor my postprocessing to the shot, as I suppose most people do I only use Lightroom as full Photoshop seems ridiculously expensive and I'm usually happy with the results I get (I do use Gimp for astrophotography, but that's all).

For birds, my primary subjects it goes like this:

Nowadays I try to expose to the right - overexposing as much as possible without blowing the highlights - at least when shooting birds, especially dark ones. I'll bring the exposure down until it looks right (I do everything by eye, to my taste), which helps keep shadow noise lower. I have always tried to use as much of the tonal range (if that's the term) as possible, I don't know why - it's just my personal style I guess. That means shadows are darkened, and highlights brightened until the full histogram space is used, with no clipping.

If I've used a high ISO (on the 5D3, 1600 and above, although it depends very much on light levels), I will apply moderate colour noise reduction, usually increasing saturation to compensate. I prefer to preserve detail at the expense of higher luminance noise, but on areas of smooth tone (e.g. out of focus backgrounds), I will often hand-paint higher noise reduction because there's no detail to lose, and this can actually improve the look of the bokeh to my eye (when done with care). With particularly difficult shots, I'll oversharpen the in focus areas and then apply wholesale noise reduction, which helps retain detail.

Since I share shots at reduced size (currently 1800 on the longest edge for Flickr where possible), I tend to split noise reduction into two parts - some on the original, then some on the resized version. At very high ISO, I tend to do several cycles, reducing the size a little each time, as it seems to provide the best results (better than all at one scale). If I'm printing large, I will reprocess, usually leaving in more noise and adding a little extra sharpness, but that varies.

Some examples. The siskin and dipper were in poor light, ISO 1600 and 2000 respectively, whereas the linnet was in ideal conditions, ISO 400 (all 1000mm, f/10).

 

[PicaPica]

I'm bored and frustrated by the never ending discussions of shadow noise, banding, etc. etc.

and that´s why you start yet another thread? ;D

i use LR noisereduction for RAW and then noiseware in photoshop.

i usually mask areas with more details.
out of focus areas get more noise reduction than high detail areas.

 

[dpc]

I generally don't worry about noise reduction. I usually use DxO Optics Pro 9 Elite (I also use Lightroom 5) and find the automatic noise reduction works very well. I have Nik Dfine 2 but almost never use it.

 

[dtaylor]

I don't actually rate Lr/ACR as highly as I used to in terms of noise handling, and give the nod to Capture One these days.

Keith - have you noticed any resolution/fine detail differences between Capture One and ACR? I only ask because the thing that made me an ACR fan was that I noticed a substantial fine detail gain with my 7D over DPP. (Not sure if that's still the case.)

For some reason I had it in my head that Capture One was expensive, but it's not. I'm going to have to give it a try.

 

[Keith_Reeder]

It's not expensive at the moment, but it was more than twice the price of Lr until recently (for the Pro version, anyway - but that's closest in functional terms to Lr).

I have to say yes to the resolution/fine detail question: there's not a gulf between them, and I'm well aware of the dangers of making decisions based on default settings; but Capture One definitely renders more micro-detail for a given level of noise, by which I mean that I can compensate in Lr for the slightly softer default rendering (which I put down to the different demosaicing algorithms in use) by moving the Detail slider along; but doing so will introduce artefacts which I don't see in Capture One for a given perceived level of detail.

I'll also add though, that in my view there's something broken in the some of Lightroom's newer camera "profiles": I've reported about the 70D, and others have made similar observations about the 6D, that the files are way noisier than they should be regardless of ISO (and they're downright grim at high ISO) in a way which no amount of additional Lightroom NR is able to fix - and you're killing detail in the attempt.

If I didn't know better*, I'd swear that these bodies are stuck on Process Version 2003 (the grotty, blotchy, noisy original rendering, which turned me off Lr for years) regardless of the selected PV.




* And, errmmmm... I don't.

 

[unfocused]

So, for those who use a RAW converter other than ACR or Lightroom, how do you get it into Photoshop after you've made your adjustments in, say, Capture One? I assume you save it as a new file format?

Sorry, but I'm not familiar with other programs and have only used Adobe.

 

[Keith_Reeder]

So, for those who use a RAW converter other than ACR or Lightroom, how do you get it into Photoshop after you've made your adjustments in, say, Capture One? I assume you save it as a new file format?

Sorry, but I'm not familiar with other programs and have only used Adobe.

Most of them have a "Send to..." option for converted files, but I've always converted and saved all of my images in one run, and then I have a second session where I work on the files (16-bit tiffs) in Photoshop.

I use this approach simply because if I've had a good day with the camera I might come back with 1,000 or more files (it's usually a lot less than that, but regularly in the hundreds - goes with the territory, shooting wildlife/sports/motor sport/air shows) and I don't want to be diving in and out of Photoshop on a file-by-file basis.

So usually it's conversions one day, post processing the next - I'm not a pro, so I'm rarely "on the clock".

 

[raptor3x]

MagicLantern DualISO.

 

[Jonasyphoto]

Try Capture1. It does a much better job on RAW conversation than Adobes. The problems you described and even the sharpness can be improve by it.

 

[Keith_Reeder]

Try Capture1. It does a much better job on RAW conversation than Adobes. The problems you described and even the sharpness can be improve by it.

Who are you replying to, Jonas? We've been discussing Capture One throughout the whole thread.

 

[dtaylor]

Thank you for your comments Keith. I'm going to give C1 a try.

 

[Keith_Reeder]

I think you'll appreciate it - and Phase One gives a very generous 60 day trial period, so you can really shake it down.

 

[RLPhoto]

ETTR, grad NDs, and the basic NR techniques. ETTR being the most useful.