I will also say this is probably not true, but for a different reason. In cam noise reduction is taking place before the codec with full sensor data.-5 for me.
Why? Because if you think how hard your computer works with something like Neat Video to get rid of noise, do you think that the camera running live compression at 30p or more is going to do a better, more accurate job? No chance in hell.
I'd much rather not have detail removed due to a NR algorithm optimised in camera for speed over accuracy. The GH5 isn't noisy to begin with, and even if you go into the higher ISOs, the noise is pretty even across the image, without any real pattern noise I can see.
Noise is a complicated beast to tame, so I'd much rather have the detail and noise there for me in post, to get rid of using an algorithm optimised for accuracy instead of speed, which is tweakable to the exact scene requirements by me.
Just my 2c.
Cheers,
Paul
GH5 NR -3 vs -5?
- Thread starter filmguy123
- Start date
I suggest that GH5 owners do their own tests using their own shooting scenarios to determine what level of sharpness and NR they are happy with after all it's not hard to do. I'd also suggest people send in a 'feature request' to Panasonic to option -5 as camera sharpening and NR turned completely off although I can see technical reasons why it may struggle with compressing all that extra 'noise data' into the bitrate limits. Maybe they would just offer it on the HDMI output?
Last edited:
I made a test. I shooted iso3200 very dark clips from NR-5 to NR +5. GH5 NR is quite subtle in range -5 to 0. GH4 has much muddier NR even at NR -4. It quite safe to use GH5 NR -5 -4 -3 -2 1- 0.
GH5 has so good codecs that it preserves the original noise very well. Seems that I cant load my example image here in DVX (maybe too big) but it is here:
https://www.dpreview.com/forums/post/60399343
GH5 has so good codecs that it preserves the original noise very well. Seems that I cant load my example image here in DVX (maybe too big) but it is here:
https://www.dpreview.com/forums/post/60399343
Last edited:
Thomas Smet
Veteran
I think some people misunderstood my first post in here. It wasn't meant to be sarcastic but really there is no better way to understand which NR setting to use then for each person to try it out for themselves.
There is a darn good reason why the GH5 give you a choice of how much to use and that is because there is no correct answer. We have choices because much like a box of crayons it is up to us to pick the one that works the best for us or a particular situation.
Best in what circumstance? HD, UHD, Opengate, 150mbps, 400mbps, external? What ISO level? Noise isn't a static sort of thing that a one size fits all solution will work for. Different ISO levels could yield very different results when it comes to NR.
There is a darn good reason why the GH5 give you a choice of how much to use and that is because there is no correct answer. We have choices because much like a box of crayons it is up to us to pick the one that works the best for us or a particular situation.
Best in what circumstance? HD, UHD, Opengate, 150mbps, 400mbps, external? What ISO level? Noise isn't a static sort of thing that a one size fits all solution will work for. Different ISO levels could yield very different results when it comes to NR.
Thomas Smet
Veteran
But you are assuming the compression can't handle the noise which may or may not be the case. Despite when the NR takes place that doesn't change the fact that the processors in the camera are already handling de-bayering or raw data, processing of the profiles and doing some pretty intensive scaling processing to get from the full sensor down to HD or UHD. Even if the processors were only doing NR they would still be greatly underpowered compared to a computer.
I think it really does depend on what ISO level one is using and what the overall exposure of the scene is like. ISO is not the only factor when it comes to noise and you can have noise at lower ISO levels or very little noise at higher ISO levels depending on the exposure of the scene.
Temporal NR may help a bit but there are so many other factors at plat which is why I said it is important for each person to test for themselves.
Thomas Smet
Veteran
I'm not sure it works as efficiently as you think it does. I'm sure there is a certain amount of ingenuity going on in the camera but I have yet to see a camera handle NR in a way superior to that in post. Efficiency or not there just isn't a ton of processing power there to use. The reality is we are all just stating theories and nothing that is cold hard facts. Thats why I said it is important for each person to make the judgement on their own. NR and noise in general is a subjective taste and much like color grading is really up to each individual. The science behind noise handling is pretty darn complex to the point where I think we are all splitting hairs guessing at answers we will likely never know or that most people will not care about. Noise reducing in general is a destructive process no matter how it is done. The aim with professional video is to never get to the situation where we even need to discuss NR. That is largely why I have not set to -5. I tend to not shoot in a way that requires any level of NR. Even if visually there doesn't seem like much of a difference between -5 and 0 the reality is there is some processing taking place on the raw pixels and I prefer to reduce that as much as possible. People wonder why the GH4/GH5 has such a video look to it and I can tell you things like NR and sharpness are some of the biggest contributing factors to that. Those are I camera attributes that don't really exist in the cinema camera world. Yes those cameras have the benefit of raw or ProRes recording which is one of the reasons why I prefer external recording at times.
stoneinapond
Veteran
Not sure how you could test that.
My point is that noise is coming off of the sensor and the corresponding A/D circuits, not the light from outside. And I'm pretty sure the engineers at Panasonic know what that level is (if they don't we're in a real heap of trouble). So they have a head start over external noise reduction software.
As you say, we are all just stating theories, but experience and logic is what is contributing to my thinking, for what it's worth.....
I tested it and looks that from NR -5 to NR 0 is very close. I also found that when using my graphics card NR when playing the files (Nvidia) it makes cleaner result in real time than camera NR.
Thomas Smet
Veteran
Which is exactly why I stand by this being a subjective topic and one with no definitive answer. There are likely as many opinions on noise and NR as there are posts on what is the best lens for camera X posts on this forum. We are all right and wrong. We have the ability to adjust NR because it is up to each individual to determine what settings work best for them.
I use -5 but I also do a lot of external recording and in that situTion I think it does make more sense to do it in post. If I was shooting a live event it would be insane to process noise in post. Just too many variables at play here to give a one size fits all solution.
...But camera NR has to happen in real-time, where NR in post can take as much time as it needs to clean up the footage. ( i.e. multiple passes with significantly more processing power than a camera has )
stoneinapond
Veteran
To noise reduce or not noise reduce, that is the question.......
You all make valid points.
As far as I'm concerned it's all freaking magic anyway. That we get the images we do from these ridiculously small instruments, is no small marvel. I'm grateful, love shooting, don't chimp, and am happiest at my editing station, NR not withstanding. :happy:
joe12south
Veteran
1. NR is *always* happening in camera. It is not possible to get a usable signal from a bayer-pattern CMOS sensor without it.
2. There is always a best time in a signal processing chain to do a given transform (chain of operation.) There is almost no operation that is best done *after* compression.
3. Comparing a purpose built ASIC to a general purpose processor (even a vector based processor like a GPU) is comparing apples and oranges.
Post-processing can do wonderful things. It can save otherwise unusable shots. I love having both NEAT and Resolve in my toolkit. BUT, as a rule...given a sufficiently sophisticated method (and the GH5's is quite sophisticated for its time)...in camera NR is better positioned to give optimal results.
With the GH5's addition of temporal processing, it is clearly, visibly superior to the GH4. I've been hard pressed to find a shot in which I would have benefited from setting it to -5 and using NEAT, vs leaving it at the default and using NEAT. I just have not seen it lose actual, usable image detail that could be saved via post-processing. (If you have an example, I'd love to see it.) I'm not saying I don't have instances where I need to use post-processing - I do for almost all footage above 1600 ISO - I'm just saying that NR -5 + post does not generate better results than NR 0 + post.
2. There is always a best time in a signal processing chain to do a given transform (chain of operation.) There is almost no operation that is best done *after* compression.
3. Comparing a purpose built ASIC to a general purpose processor (even a vector based processor like a GPU) is comparing apples and oranges.
Post-processing can do wonderful things. It can save otherwise unusable shots. I love having both NEAT and Resolve in my toolkit. BUT, as a rule...given a sufficiently sophisticated method (and the GH5's is quite sophisticated for its time)...in camera NR is better positioned to give optimal results.
With the GH5's addition of temporal processing, it is clearly, visibly superior to the GH4. I've been hard pressed to find a shot in which I would have benefited from setting it to -5 and using NEAT, vs leaving it at the default and using NEAT. I just have not seen it lose actual, usable image detail that could be saved via post-processing. (If you have an example, I'd love to see it.) I'm not saying I don't have instances where I need to use post-processing - I do for almost all footage above 1600 ISO - I'm just saying that NR -5 + post does not generate better results than NR 0 + post.
In my tests GH5 NR-5 and graphics card real time NR gives more pleasing result than any GH5 NR setting 0, +2, +5 etc. GH5 NR -5 has already good noise reduction from original signal. But yes GH5 NR is subtle and it is not making so muddy result than GH4. GH4 NR -4 is muddier than GH5 NR +1.
I'm not convinced because this was NOT the case for the GH4 camera based upon my own tests and 2+ years of shooting with it.
I haven't compared GH5 NR with NEAT NR, but with the GH4 you could retain better image detail by shooting with camera NR at -5 and then applying calibrated NEAT NR in post.
( heavy camera NR would often "eat" the fine details that were retained by doing NR in post )
When I can find some time I will test this to see if things have changed with the GH5.
joe12south
Veteran
GH4 <> GH5 noise reduction is quite different. Check it out.
joe12south
Veteran
What does "more pleasing" mean? It would be more useful to frame the discussion around something empirical. My measure of that is which method preserves the most "signal" (image detail) while removing the most noise.
What I think you'll find (at least I did) is that if you dial down the GH5's NR, there is no meaningfully additional "signal" to be found in the noisy areas for a post-processing solution to work with. All the post-processing can do is further smooth those noisy areas. Those areas become more plastic vs revealing texture.
GH5 NR blurs the image. The blurring is subtle. There is not much difference between NR -5 and NR 0. My "empirical" experience is that the graphics card NVIDIA NR is more clever to hide visible noise and maintain details than GH5 NR. I speak about playing video and what I am seeing in my big 65" TV as a computer monitor. Nvidia NR makes more than just blurs the video. It "slows" the noise action or noise frame rate somehow so that it is not so "active". It is hard to explain but the noise looks calmer or not so nervous. i think it also blends some tones so that there may be little more banding in some parts of image in max setting.
At low iso (100-200) GH5 has so little noise that even NR -5 is practically noiseless.
joe12south
Veteran
What's the point of Nvidia's NR (or EE) for content creation? It happens at the point of display, so unless you're sending your graphics card to each viewer, it's not a factor. ;-)
Regardless, if it really does feature an algorithm that retains more detail, I would genuinely like to see a screen grab, because I would consider NEAT Video amongst the best.
PS. If you underexpose the GH5, even at ISO200, you'll get a lot of noise. Important to note, because better to use a higher ISO and properly expose than force a lower ISO hoping for less noise.
-I mean that GH5 NR is not the best if a player or graphics card can make more pleasing NR in real time. Neat Video is probably even better. You can easily test player/card NR yourself. We are talking about quite marginal differences. I am happy that GH5 NR is not adding mud like GH4 does.
-Yes it is some times a tough decision to choose iso.