Low ISO vs. Denoise + High ISO - Watch This Before You Next Shoot!

[Steve]

In this video, I challenge the statement, "I don't need to worry about high ISO, I have denoise software!"

To what extent is that true? When do you start losing detail? When does AI software start making up detail? Can denoise software really fix everything? Checkout the these tests using Lightroom, DxO, and Topaz for the answers!

We'll also talk about getting better results with high ISO, my favorite guideline for ISO, my software workflow order, and field techniques to improve your chances with high ISO.

Grab a cup of coffee and settle in - there's a TON of critical info in this one - don't take another photo until you see it!

 

[Cameron T]

Without watching it yet, I'm going to guess the result of low iso (while still having a proper shutter speed to eliminate motion blur) is better than high iso plus denoise.

But a high iso plus denoise shot is better than a blurry one, or none at all.

 

[JAJohnson]

In the northern hemisphere, daylight hours are getting shorter and I don’t have to be up and out quite so early nor wait quite so long for sundown. I seem to do more high ISO shooting mid autumn thru mid spring.

For me this video is timely. As always, Thanks Steve!

Addendum: My usual upper limit on ISO is 12,800 but I try to stay at 8,000 or less if I can. For denoise I use Lr/LrC or Topaz Photo AI. I’ll probably take a look at DxO in the coming weeks.

 

[Cameron T]

In the middle of watching the video now, I stopped using topaz because for most of my shots because it kept doing weird things to them, and I use dxo pure raw now.

My upper limit is set to 6400, and I'm happy with that for birds. If I'm shooting other things, I might up it some (or if I'm on some unique trip, etc), but for day to day it's a good check on "do I need to lower shutter speed, do I really want to shoot in this light".

As mentioned, it's also depending on where you post and what you're after as the end result.

For most of my images, I export at 50 or 75% for web use, so I've been happy with dxo.

 

[Steve]

Yeah, I really like DxO for the speed. I don't use it every time (I don't really use it the majority of the time I suppose), but it's quick and when it works, it saves a lot of time

 

[Steve]

Oh, here's that Lightroom Denoise video I was talking about in the main video above, if anyone is interested:

 

[Cameron T]

Yeah, I really like DxO for the speed. I don't use it every time (I don't really use it the majority of the time I suppose), but it's quick and when it works, it saves a lot of time

I've found turning the sharpening down from pure raw worked well for me and most of my images. It stops the over sharpening, while with topaz (for me) it never seemed to work as well as I'd like.

To be fair, all of my experience is with the newest denoise models (the last 3 iterations), while topaz was mostly the older ones.

 

[DsD]

Great vid Steve. Yes the struggle is real!

I just came back from Montana and most of my shots were with the 800 pf on the Z9 and between what I think were atmospherics, not being able to fill the frame adequately and motion blur while trying to keep ISO in check I think most all my shots are junk! Sometimes I think one needs to know when to "fold em" rather than trying to "save" shots in less than optimal conditions.

The rancher that hosted us said he wanted to get some of my photos but my brother warned him that I am like that kid form the 70"s cereal commercial ... "Mikey hates everything"! I believe I am in the super critical camp concerning retaining detail. I did unknowingly employ your great tip of defocusing and refocusing , both with and without subject detection, to acquire sharp focus but fear I failed most times. Actually, I am quite demoralized after that trip.

Per your request, I only use DXO as my developer but do see the advantage of using different denoise programs now that you did the comparisons.

Thanks for the ISO confirmation education!

Derek

 

[Gaduf]

Great video Steve,
Thanks for sharing!
Gavin

 

[JAS]

Well done and helpful in figuring things out when shooting and high ISO is a driver. With shaky hands and physical limitations, I am usually at a higher ISO than I might prefer, so I rely on the software to help. My job is made easier in that most of my work is for online posting, 12x12 or 11x14 photobooks and prints not larger than 13x19. So, I have a bit more latitude in what is acceptable to me in my circumstances. I do prefer LrC Denoise and use it for most of my shots needing more than minor adjustment. DxO PL8 is my second choice for when doing my IR photography because it meshes better with Nik Suite for the B&W conversions.
Thank you and looking forward to your next nuggets of wisdom!

 

[Woodyg3]

I appreciate all the work you put into this. A friend of mine has been trying to get me to give DXO a go, but from your examples I really like Lightroom the best. Others may come to a different conclusion, of course. For me, Topaz has been relegated to "saving" a picture will a little bit of motion blur, and when I say "saving" I mean doing up with something I can share online, not something I would print of have displayed in a large format. I think Lightroom has surpassed Topaz in removing noise, and I don't use Topaz for that purpose any more.

When shooting owls before sunrise or after sunset, I've taken plenty of shots at 12,800. One of them is hanging on the wall in my house. At a certain point you just have to do the best you can in a given situation. That said, I try really hard not to go beyond 4-5,000 ISO.

 

[ajrmd]

Very interesting video and it seems to confirm my observations albeit I use different NR tools to achieve the end results. What I find very useful is using iterative applications of targeted NR/Sharpening are more important than one realizes. Most default NR is heavy handed and destroys detail while introducing artifact. Likewise, there is little one can do regarding the loss of color and image fidelity when using high ISO's other than bracketing and pushing the envelope to reduce ISO when feasible. I should add that no NR software has ever "magically" saved an image making it suddenly suitable for printing. Yes, it has saved a few images for web use, but since it can't produce real detail and color fidelity, this doesn't translate into printing.

 

[RichF]

In this video, I challenge the statement, "I don't need to worry about high ISO, I have denoise software!"

To what extent is that true? When do you start losing detail? When does AI software start making up detail? Can denoise software really fix everything? Checkout the these tests using Lightroom, DxO, and Topaz for the answers!

We'll also talk about getting better results with high ISO, my favorite guideline for ISO, my software workflow order, and field techniques to improve your chances with high ISO.

Grab a cup of coffee and settle in - there's a TON of critical info in this one - don't take another photo until you see it!

Looking forward to watching this. As always, I suspect Steve's video will be excellent.

 

[macwalter]

This is for me a Complicated subject. First off I handhold and realize the limits of this technique. Also I am not always able to get close enough to fill the frame. Recently I having moved up to the 600pf and even with the 1.4tc has allowed me to get larger images compared to the 400f4.5 with the 1.4tc. The 600pf gives cleaner more detailed images in most lighting conditions. In very low light the 400 is my lens of choice.
in spite of this I am pretty obsessed with feather detail and try to get as much as I can.
so far I havent had as much success with LR Denoise as I don’t have as much control. I find myself tending to prefer DxO overall and I use the advanced settings a lot to try to compromise on real detail and Ai. I don’t like smooth featureless feathers. I find I can often dial in a decent compromise with DxO. I also like that it is fast. I also have topaz photo aI and use it at times. I find with low to moderate iso say 3200 or lower it can do an excellent job but at high iso I find it very uneven and there is not enough control to modify this. As Steve said the sharpening is very aggressive and most of the models are worthless. So I prefer DxO overall. I need to experiment with LR more I guess.
In the field I try to keep iso below 4000 but I have had good results as high as 12800 in good light with a close subject.
(By the way I hate typing in this software )

 

[RichF]

Looking forward to watching this. As always, I suspect Steve's video will be excellent.

Just watched the video. Meet my expectation - a fabuous video.

 

[ButlerKid]

This is for me a Complicated subject. First off I handhold and realize the limits of this technique. Also I am not always able to get close enough to fill the frame. Recently I having moved up to the 600pf and even with the 1.4tc has allowed me to get larger images compared to the 400f4.5 with the 1.4tc. The 600pf gives cleaner more detailed images in most lighting conditions. In very low light the 400 is my lens of choice.
in spite of this I am pretty obsessed with feather detail and try to get as much as I can.
so far I havent had as much success with LR Denoise as I don’t have as much control. I find myself tending to prefer DxO overall and I use the advanced settings a lot to try to compromise on real detail and Ai. I don’t like smooth featureless feathers. I find I can often dial in a decent compromise with DxO. I also like that it is fast. I also have topaz photo aI and use it at times. I find with low to moderate iso say 3200 or lower it can do an excellent job but at high iso I find it very uneven and there is not enough control to modify this. As Steve said the sharpening is very aggressive and most of the models are worthless. So I prefer DxO overall. I need to experiment with LR more I guess.
In the field I try to keep iso below 4000 but I have had good results as high as 12800 in good light with a close subject.
(By the way I hate typing in this software )

Have you tried the de-noising process Steve recommends in THIS video?

Make ISO 12,800 Look Like ISO 400: A Lightroom Denoise Master Class - Backcountry Gallery

A Master Class On Lightroom Denoise! This video is a master class on Lightroom Denoise! If you want to see my techniques for making ISO 12,800 look like ISO 400, you’re gonna love this! It’s the BEST way to fix a noisy photo! In fact, I think the advanced techniques at the end of the […]

backcountrygallery.com

 

[DHL]

In this video, I challenge the statement, "I don't need to worry about high ISO, I have denoise software!"

To what extent is that true? When do you start losing detail? When does AI software start making up detail? Can denoise software really fix everything? Checkout the these tests using Lightroom, DxO, and Topaz for the answers!

We'll also talk about getting better results with high ISO, my favorite guideline for ISO, my software workflow order, and field techniques to improve your chances with high ISO.

Grab a cup of coffee and settle in - there's a TON of critical info in this one - don't take another photo until you see it!

Excellent video.

This really amplifies the flexibility that comes with fast glass. I recently posted a photo of a brown bear captured in the early morning at Lake Clark this past July using a Nikon Z400 f2.8 VR TC S at 400mm, ISO 2500, f2.8, 1/1000. With my 100-400mm at f5.6, all other things equal, the ISO would have been 10,000. With a 400mm f4.5, the ISO comes down, but only to 6400. Of course, I could have lowered my shutter speed if I had been using slower glass, but the bear was chewing and it likely would have impacted sharpness.

 

[macwalter]

One thing I forgot to mention is that at times I will when having processed a photo in DxO I will in LR use the brush to highlight areas where I didn’t care for the sharpening and use the sharpening slider or the texture slider to pull out a bit of the detail to dull the effect a bit. A little goes a long way

 

[Anjin San]

Couple of thoughts…especially as I’m one of the people that say we have good NR software these days…but no rage though. Obviously the tips to shoot as low as possible and fill the frame are good…but my comments about the NR software being good aren’t in the vein of “we don’t need to worry about high ISO” but rather in the threads where I say that in talking about faster lenses and the difference they make…of course they make a difference but my point was that for many of us the output usage and other factors mean that the NR is good enough. Steve has a different need than most of us that aren’t making our living with photos…he is and for his purposes the exotic expensive lenses and getting absolutely best IQ are what pays his bills…while the rest of us amateurs don’t necessarily have the same requirements. I haven’t ever said I don’t care about high ISO…shoot as low as possible but in situations where it is needed then go ahead. For something like the GBH…maybe I skip the shot because I’ve got thousands of GBH shots…but for the pair of lions drinking by the stream at suspense…yeah, I’m taking those. I was at 840mm for those shots with my 600PF and the bean bag and long bursts let me get usable for me shots…whether Steve would agree they’re usable is a different story but he’s been to Africa a lot and has had way more lions drinking at suspense opportunities than I have. You do what you can to get low ISO and critical sharpness…but then NR helps…not as much maybe as a faster lens and NR but if one doesn’t have the 600TC lens for several reasons then you get the best shot you can and PP. Steve needs to be picky…the rest of us don’t need to but would want to be as picky…but then all of those other factors come into play in determining how picky am I willing to be and at what cost “money, capability, physical limitations, amount of practice and the host of others). Like I said about the lion drinking…nobody that looked at the blog commented and said it would have been a better shot at ISO 500 instead of whatever it was. They said…cool photo. There are some even in the non pro category that demand the best and are super picky…and theres nothing wrong with that.

Second…where the output is going makes a difference…print or magazine sales require a different level of “good” than travel blog or amateur portfolio shots.

Thirdly…and few discussions of the high ISO issue get into this…pixel peeping at 2:1 isn’t necessarily the same as looking at final output images at the size they will be used. Yes, peeping is necessary while working on shots…but nobody ever does these sort of comparisons at output. Downsampling for print can’t help but reduce fine detail…and for screen even more reduction happens, that’s just physics for you. That loss means that the gap between the fully processed images gets less…and the better is the enemy of good enough idea comes into play. I’m not saying peeping isn’t needed and good…but final output comparisons at the size they’ll be used comes into play as well.

All that said…another excellent video showing why one needs to get ISO as low as feasible…and a good comparison of both the various NR methods and that one isn’t always better.

I have both DxO and Topaz…and almost all my NR is in DxO because I can batch process the selects and come back later…my patience level and LR skill level make it generally do better for me than using his manual LR method. I agree that Topaz oversharpens and DxO is better at either toning it down or leaving it off and sharpen in LR because I’m better at sharpening there than in NR. The differences he was pointing out in the details weren’t really very apparent when I looked at it on my iPad…some of that is probably YouTube compression but I’m going to look at it again on my Studio Display and see if they’re more apparent. On the iPad seeing the NR is easy but the details differences and stringiness/fake detail comparisons were hard to see…hopefully the better monitor will show them better.

 

[Anjin San]

Excellent video.

This really amplifies the flexibility that comes with fast glass. I recently posted a photo of a brown bear captured in the early morning at Lake Clark this past July using a Nikon Z400 f2.8 VR TC S at 400mm, ISO 2500, f2.8, 1/1000. With my 100-400mm at f5.6, all other things equal, the ISO would have been 10,000. With a 400mm f4.5, the ISO comes down, but only to 6400. Of course, I could have lowered my shutter speed if I had been using slower glass, but the bear was chewing and it likely would have impacted sharpness.

I’m not going to say fast glass isn’t better because from a purely scientific viewpoint it is better…but IMO the advantage while it does exist is a bit over amplified. The 600TC with the TC in becomes 840/f5.6 for instance and the 600PF becomes 840/f9…so a stop and a third slower. Yes, that’s a higher ISO…but the video indicates that 2-3 stops can be reasonably overcome (depending on your level of pickiness) so one and a third fits the gap. Given that…fast glass isn’t necessarily the panacea that it sometimes gets blessed with…sure, it is better but as I said in the other longer reply better is relative and that faster glass comes with other drawbacks, cost is a big one but even if one ignores money cost there are other costs considered in making a lens selection that may or may not make faster glass the right choice.

 

[Calson]

Periodically I review the various NR applications to see which one is currently the best at cleaning up noise with minimal loss of detail and color fidelity. I use a woman's face and use the skin tones to evaluate color fidelity. I have seen the same with images of the throat area of male sea lions where the throat skin has a good deal of tonal and color variation.

NR apps often compress the tonal range and I find this to be as much of a problem as the loss of detail in images. "Noise" is the effect when a sensor is not able to capture image details properly. NR software may reduce the appearance of noise but always with a loss of detain and color fidelity. Wiith some subjects this does not matter but for most it is quite apparent.

 

[Steve]

The differences he was pointing out in the details weren’t really very apparent when I looked at it on my iPad…some of that is probably YouTube compression but I’m going to look at it again on my Studio Display and see if they’re more apparent. On the iPad seeing the NR is easy but the details differences and stringiness/fake detail comparisons were hard to see…hopefully the better monitor will show them better.

I did it in 4K, but YT is really compressing it. I've had another person tell me that the compression was making it tough to tell.

 

[Ralph]

Yeah, I really like DxO for the speed. I don't use it every time (I don't really use it the majority of the time I suppose), but it's quick and when it works, it saves a lot of time

My upper limit on my D500 is 4000. I use DXO as well when needed.

 

[macwalter]

One other thing. DxO has just released a new Denoise algorithm and I do find a slight improvement. It seems easier to tame excess sharpening.

 

[macwalter]

Couple of thoughts…especially as I’m one of the people that say we have good NR software these days…but no rage though. Obviously the tips to shoot as low as possible and fill the frame are good…but my comments about the NR software being good aren’t in the vein of “we don’t need to worry about high ISO” but rather in the threads where I say that in talking about faster lenses and the difference they make…of course they make a difference but my point was that for many of us the output usage and other factors mean that the NR is good enough. Steve has a different need than most of us that aren’t making our living with photos…he is and for his purposes the exotic expensive lenses and getting absolutely best IQ are what pays his bills…while the rest of us amateurs don’t necessarily have the same requirements. I haven’t ever said I don’t care about high ISO…shoot as low as possible but in situations where it is needed then go ahead. For something like the GBH…maybe I skip the shot because I’ve got thousands of GBH shots…but for the pair of lions drinking by the stream at suspense…yeah, I’m taking those. I was at 840mm for those shots with my 600PF and the bean bag and long bursts let me get usable for me shots…whether Steve would agree they’re usable is a different story but he’s been to Africa a lot and has had way more lions drinking at suspense opportunities than I have. You do what you can to get low ISO and critical sharpness…but then NR helps…not as much maybe as a faster lens and NR but if one doesn’t have the 600TC lens for several reasons then you get the best shot you can and PP. Steve needs to be picky…the rest of us don’t need to but would want to be as picky…but then all of those other factors come into play in determining how picky am I willing to be and at what cost “money, capability, physical limitations, amount of practice and the host of others). Like I said about the lion drinking…nobody that looked at the blog commented and said it would have been a better shot at ISO 500 instead of whatever it was. They said…cool photo. There are some even in the non pro category that demand the best and are super picky…and theres nothing wrong with that.

Second…where the output is going makes a difference…print or magazine sales require a different level of “good” than travel blog or amateur portfolio shots.

Thirdly…and few discussions of the high ISO issue get into this…pixel peeping at 2:1 isn’t necessarily the same as looking at final output images at the size they will be used. Yes, peeping is necessary while working on shots…but nobody ever does these sort of comparisons at output. Downsampling for print can’t help but reduce fine detail…and for screen even more reduction happens, that’s just physics for you. That loss means that the gap between the fully processed images gets less…and the better is the enemy of good enough idea comes into play. I’m not saying peeping isn’t needed and good…but final output comparisons at the size they’ll be used comes into play as well.

All that said…another excellent video showing why one needs to get ISO as low as feasible…and a good comparison of both the various NR methods and that one isn’t always better.

I have both DxO and Topaz…and almost all my NR is in DxO because I can batch process the selects and come back later…my patience level and LR skill level make it generally do better for me than using his manual LR method. I agree that Topaz oversharpens and DxO is better at either toning it down or leaving it off and sharpen in LR because I’m better at sharpening there than in NR. The differences he was pointing out in the details weren’t really very apparent when I looked at it on my iPad…some of that is probably YouTube compression but I’m going to look at it again on my Studio Display and see if they’re more apparent. On the iPad seeing the NR is easy but the details differences and stringiness/fake detail comparisons were hard to see…hopefully the better monitor will show them better.

I’m with you on this. I mainly pixel peep to compare 2 or more very similar images to find the one with the best detail. Then I will zoom out and think about how I will crop etc. dox works best for me due to the batch processing and I can go through a large number of shots fast and tweak them in DxO for a bit better result. Also your right about comparing us normal people who can’t go to Africa but maybe a couple times in our lives vs those that act as guides etc. or are extremely wealthy who go every year and to the best places with the most opportunities.