A General discussion of noise and what might be done

[vindibona1]

We all know that low light conditions often, if not usually bring noise. We know that the current and upcoming DJI drones have address low light and noise in various ways. The two prong attack that DJI (and Autel) have implemented is a larger sensor with a wider F1.7 aperture lens. Big improvements out of the box, for sure. But to avail ourselves to those improvements involves some costs, some greater than others. I would like to suggest that, while not as simple as just buying an upgraded drone, if noise is what's driving you, perhaps there are alternate methods of noise reduction with a better ROI where you might reconsider upgrading your drone for this purpose. I don't know.

Some time ago, even before the Air2s was released I did some experimentation, in Davinci Resolve doing noise reduction on its own, and with Davinci using the Neat Video Plugin. I would like to offer samples of these video clips to see what you think, perhaps causing some rethinking about drone upgrades. Again, I don't know. It's up to you. I'm just putting it out there. Two short clips here, if nothing else to provide context to low light noise. BTW... The Bahai Temple (Wilmette IL) was taken before 5am.

 

[MS Coast]

Nice work. Interesting. Informative.

Thanks for taking the time to present your results. Actual results are so much more useful than assertive opinions.

 

[SethB]

Neat Video is a very cool product, and very worthwhile if you need it!

Low light performance is a leading driver in *camera* manufacturers’ new product releases. The tech evolves rapidly. It’s not only sensor size; all modern cameras are doing noise reduction in-camera prior to encoding.

I don’t have insider information as to where NR fits in DJI’s product marketing/development plans, but many gains can be found in firmware releases that include new noise reduction methods.

 

[vindibona1]

Nice work. Interesting. Informative.

Thanks for taking the time to present your results. Actual results are so much more useful than assertive opinions.

You are most welcome.
There are a couple other takeaways that might be helpful, especially when there is a debate (at least in my mind) about which drone provides the best resolution and detail...

Neat Video has two levels of pricing; The standard which only works up to 1080p which cost me $90 and the and the pro, with no res limit size of $160. I didn't know what I was doing but didn't want to spend $160 for an experiment. so went with the $90 version and applied it to a 1080p clip, then upscaled. It was shot in 4k and then downscaled to 1080p for noise removal, then upscaled again. In other words, interpolated twice. I wasn't aware of the noise removal capabilities of DR17 either. So, if you look at the last example with Neat video (basic) I took a 1080p file and upscaled it to 4k. While I'm sure it wouldn't be the same detail and res as an original 4k file, but it really isn't bad. But again, it's on YouTube which has it's own compression.

This again brings me to conclude that it's not just the resolution and detail of the original file that counts, but the impact of how the the viewers get to look at the files combined with the equipment they are viewing it with. One other thing, most video clips are often very limited in clip length, especially B roll. In fact, when I'm watching a lot of videos I'm frustrated at how quickly many of them jump from scene to scene. I'm fine paying $160 to make my noisy videos acceptable, IMO a far better alternative than spending hundreds, if not thousands to upgrade my drone equipment for limited benefit. JMO

 

[MS Coast]

You are most welcome.
There are a couple other takeaways that might be helpful, especially when there is a debate (at least in my mind) about which drone provides the best resolution and detail...

Neat Video has two levels of pricing; The standard which only works up to 1080p which cost me $90 and the and the pro, with no res limit size of $160. I didn't know what I was doing but didn't want to spend $160 for an experiment. so went with the $90 version and applied it to a 1080p clip, then upscaled. It was shot in 4k and then downscaled to 1080p for noise removal, then upscaled again. In other words, interpolated twice. I wasn't aware of the noise removal capabilities of DR17 either. So, if you look at the last example with Neat video (basic) I took a 1080p file and upscaled it to 4k. While I'm sure it wouldn't be the same detail and res as an original 4k file, but it really isn't bad. But again, it's on YouTube which has it's own compression.

This again brings me to conclude that it's not just the resolution and detail of the original file that counts, but the impact of how the the viewers get to look at the files combined with the equipment they are viewing it with. One other thing, most video clips are often very limited in clip length, especially B roll. In fact, when I'm watching a lot of videos I'm frustrated at how quickly many of them jump from scene to scene. I'm fine paying $160 to make my noisy videos acceptable, IMO a far better alternative than spending hundreds, if not thousands to upgrade my drone equipment for limited benefit. JMO

"This again brings me to conclude that it's not just the resolution and detail of the original file that counts, but the impact of how the the viewers get to look at the files combined with the equipment they are viewing it with. "

Amen.

I'd guess that at least half the amateur videos made these days are viewed almost entirely on cell phones. Those made with drones are more likely to be seen on a tablet or computer but cell phones probably still dominate. A small portion are probably viewed on televisions, but I suspect that only a minority are viewed on a device that provides 2.7k or better resolution.

So, a Mini is probably adequate for most drone video makers who are doing only basic editing. There's certainly a benefit to better devices to allow more sophisticated editing, cropping, color grading, shooting under more adverse conditions, and such. I'd bet that most of us are farm more limited by our shooting and editing skills than by our hardware. I know for certain that's true for me.

But will I seriously consider a new, shinier, slicker, more powerful drone with a better camera? Oh, yeah.

 

[vindibona1]

But will I seriously consider a new, shinier, slicker, more powerful drone with a better camera? Oh, yeah.

The Mini 2 was quite a step up from the Mini 1 in many areas. When I got my Mini 1 I didn't know what droning involved and had debate on getting the MA2 as a starter, but the fugal side of me won. I thought 2.7k was adequate, and adequate would in fact describe the video quality. What bothered me was that the wind here was such that I wasn't flying quite literally 50% of the time I wanted to fly; conditions that the Mini 2 would have handled. So 4 weeks after getting the Mini 1 I was convinced that droning was a great thing for me and I bought a refurb M2P. Great decision, btw.

What the Mini 3 might offer in terms of image/video quality (in the file) would be an increased write speed which will add to the 4k quality. It's going to be a winner with a lot of folks who need that light weight drone. That's for sure. But the one thing that I experience with my Mini 2 that drives me crazy sometimes are the dead spots in the controller which makes my Mini 2 far less responsive than my M2P. I know I've harped on this earlier, but the amount of programmable button commands on the M2's controller are quite extensive, something I miss on the Mini 2, which aren't there on any of DJI's current drones, even through M3's "standard" controller.

 

[wco81]

What about for stills?

See some nice low-like photos posted here.

ISO is only a few hundred, which on a full frame camera is nothing.

But out of 1-inch sensor such as the M2P, the noise is very dominant even at a couple hundred ISO. These are RAW files too.