Mavic 2 Pro 4K HQ vs 4K FOV

[gnirtS]

Not astonishing at all assuming you tested up close given optics on the mavic are specifically optimised for longer distances.
You're testing the drone at the worst possible situation possible (and artificial as you're never going that close).
Get a bigger chart, try it from 50m distance, make sure you're using specific apertures and ISOs then try again.

 

[Mike_Jaxson]

Well this topic bothered me, therefore I made the test myself with all the cams I have. Sony AX100, Hero 5, Mavic 1 Pro, Mavic 2 Pro...giving them all the same challenge within the same test environment. The result is a little astonishing...what do you think?

View attachment 46391

That you need to include the Zoom. Seriously though they hopefully can do better with a patch. The H3 site says stuff about 8k. Ambarella | Embedded Computer Vision SoCs

 

[Sockeye]

That you need to include the Zoom. Seriously though they hopefully can do better with a patch. The H3 site says stuff about 8k. Ambarella | Embedded Computer Vision SoCs

Would have loved to, but I simply did not buy it because I hoped for a higher image quality of the M2P

 

[Mike_Jaxson]

Would have loved to, but I simply did not buy it because I hoped for a higher image quality of the M2P

You bought the best portable drone DJI makes. In the mean time go get some shots in HQ mode and wait on the patch.

 

[Sockeye]

Not astonishing at all assuming you tested up close given optics on the mavic are specifically optimised for longer distances.
You're testing the drone at the worst possible situation possible (and artificial as you're never going that close).
Get a bigger chart, try it from 50m distance, make sure you're using specific apertures and ISOs then try again.

Mhmm... The distance was around 1,5m for the FOV and around 2m for the HQ. Aperture setting f2.0 (should be the best) and ISO100.

But could you let me know where you have the information from, that the lenses of the M2P have a higher resolution when focusing objects further away?

But regardless of that. The test is just comparing HQ and FOV and more or less proves that the downscaling from the full sensor to the 4K is most probably done by pixel binning/line skipping and not calculated as the 1" Sony or the 1" P4P do.

Let's hope DJI hears us and applies a fw update soon.

 

[gnirtS]

Mhmm... The distance was around 1,5m for the FOV and around 2m for the HQ. Aperture setting f2.0 (should be the best) and ISO100.

But could you let me know where you have the information from, that the lenses of the M2P have a higher resolution when focusing objects further away?

But regardless of that. The test is just comparing HQ and FOV and more or less proves that the downscaling from the full sensor to the 4K is most probably done by pixel binning/line skipping and not calculated as the 1" Sony or the 1" P4P do.

Let's hope DJI hears us and applies a fw update soon.

An engineer involved (Ming Thien, notes on his blog) has stated the optics are optimised for distance as it typical with an aerial shot (very common with drones) not close up. The MP doesn't do f/2 aperture?

Sharpness tests show f/2.8 to 5.6 being ok with f/4 about the sharpest..

 

[Sockeye]

oops... was a typo. f2.8

 

[Sockeye]

Sharpness tests show f/2.8 to 5.6 being ok with f/4 about the sharpest..

I also saw other tests...

But I will also conduct an aperture comparison for myself. What distance would this Lens Engineer recommend?

 

[gnirtS]

He specified "distance". Given that i'd expect it to mean out past infinity, potentially more if its optimised for flight.

The other issues with the tests are no details of the sharpening settings used on mavic 1 and 2, whether it was in log or other and so on.
(im guessing these are video not stills)

 

[Sockeye]

Yes of course Video. And again just to prove the crappy down-scaling of FOV. The color-format is not relevant and sharpening set to 0.

And btw. with testing the 2.7K and 1080p modes with 60fps I found out that they also have ugly binning/line skipping issues.

 

[yysc]

It seems 4K FOV is indeed 2.7K FOV upscaled this would explain the resolution loss.

"Full FOV 4K is actually 2.7K upscaled, as the IMX183 sensor can either do full 5.5K readout or 2.7K readout, the Ambarella chip obviously cannot handle 5.5K supersampling."
DJI Mavic Pro II

From Sony
IMX183CQJ | Sony Semiconductor Solutions

If the P4P with the H1 can do proper 5.5K to 4K downsample the M2P with H3 can too. The key for me is the sensor can read 5.5K at 10bit at 25fps max.

The P4P was recording 8bit hence it could do 5.5K 30fps readout and then proper downsample.

If this turns out to be a sensor limitation DJI needs to offer us 4K FOV 8bit "HQ" and the regular upscaled 4K FOV 10bit 30fps, or just switch to 8bit only in 4K FOV 30fps mode.

 

[eukaryote]

I would be more than happy with 25fps with full readout and downsampling at 10 bit. I am fairly certain the H3 has the power to do this....

 

[msavic6]

I would be more than happy with 25fps with full readout and downsampling at 10 bit. I am fairly certain the H3 has the power to do this....

I would be happy with this as well but also would like the option to shoot 30 FPS full readout at 8 bit.

 

[MobiosOne]

Correct me if I'm wrong but what you are saying is that we can't have full sensor reading at 5.7k because the H3 can't handle it at 10bit @30 fps?
It seems kind of a waste to have a 1" sensor and either downsample the image to oblivion or just crop the image down to the size of a 4k sensor
I rather have full readout at 8 bit and 30fps than a mushy 2.7k downsample image

 

[ArchieFPV]

Great overview. I must say I'm a bit disappointed.

 

[Roamer105]

While we are arguing about detail resolution we tend to overlook the real qualities of the video image produced by the M2P. Please check out this white paper about understanding sharpness: http://www.falklumo.com/lumolabs/articles/sharpness/ImageSharpness.pdf - really good read with some good explanations for some of what we see.

Here are three frame grabs from the same subject video taken with a P4P (Dlog, style=standard) and a M2P (Dlog style=standard) in HQ mode (moved further away to achieve same framing) and an M2P in FOV Dlog (again distance adjusted for same framing). All footage had been graded to attempt a match using the same SkyGrade LUT.

If you zoom in a bit on each of these images looking at the UPS labels (especially the "P") you notice a significant contrast halo (due to over sharpening, over contrasting) on the P4P image and the same is very natural and smooth on both the HQ and the Full FOV images of the M2P. The "Halo" makes the P4P image appear more detailed and sharp - but at the cost of the "halo" artifact which causes the image to appear less natural.

 

[kilomikebravo]

Very interesting comparison Roamer, thanks for doing the test. Yep, definitely a halo on the P4P (at the P of course.). It’s not blatant but it is certainly there and frankly with my old eyes, the P in the FOV image looks the best of all. Now where are my dang cheaters?

KB

 

[Roamer105]

Very interesting comparison Roamer, thanks for doing the test. Yep, definitely a halo on the P4P (at the P of course.). It’s not blatant but it is certainly there and frankly with my old eyes, the P in the FOV image looks the best of all. Now where are my dang cheaters?

KB

Also, while it is quite easy to sharpen the M2P footage without adding any halo or such artifacts (short of adding a bit of noise if a lot of sharpening is applied), it is very difficult, next to impossible to remove the "halo" effect. Softening does nothing useful short of minimizing noise.

 

[czajunia]

I am trying to find a comparison of a Mavic Pro's 4K HQ Mode vs Zoom at 40/48mm. Just to see whether the cropped/zoomed image of a larger sensor offers similar quality to optical zoom of a smaller one in terms of resolution and general picture quality. I know that with the Pro we don't get all the benefits Mavic Zoom has to offer including smooth optical zoom to 48mm but that's not relevant to me here. I just wonder if both cameras can resolve similar amount of detail and offer similar picture quality when zoomed in (or cropped to HQ in case of the Pro).

Has anyone come across such a comparison? Cheers.

 

[tomibeg]

I am trying to find a comparison of a Mavic Pro's 4K HQ Mode vs Zoom at 40/48mm. Just to see whether the cropped/zoomed image of a larger sensor offers similar quality to optical zoom of a smaller one in terms of resolution and general picture quality. I know that with the Pro we don't get all the benefits Mavic Zoom has to offer including smooth optical zoom to 48mm but that's not relevant to me here. I just wonder if both cameras can resolve similar amount of detail and offer similar picture quality when zoomed in (or cropped to HQ in case of the Pro).

Has anyone come across such a comparison? Cheers.

I had exact the same question. And for my research - M2Pro in HQ mode has a bit more detailed image than M2Zoom at 40mm. But if you want to compare M2Zoom at 48mm, you have to digitally scale M2Pro HQ footage by additional 20%. And in this case M2Zoom becomes a bit cleaner, since it's completely optical, but still very similar with scaled M2Pro. All depends from digital scaling algorithm you applied to M2Pro. But again, the level of details and sharpness are VERY similar even after 20% digital scaling. And if you're ok with 40mm instead of 48mm, then you can use HQ mode without additional digital scaling, and it will give you sharper results than you could achieve with M2Zoom at 40mm.

I picked M2Pro, because I don't need to zoom while I filming, and HQ mode gives the same "cinematic" perspective compression (so it's not an exclusive "feature" of M2Zoom), also we gain 10 bit, D-Log and much better low light capabilities (thanks to constant f/2.8 aperture and 1 inch sensor)

You can find good comparison between M2Z 48mm and M2P HQ mode in this article: DJI Mavic 2 Pro VS Mavic 2 Zoom: The Ultimate Comparison | MyFirstDrone