Please select Into the mobile phone version | Continue to access the computer ver.
Why is 4K Full FOV (28mm) videomode so soft?
990 22 2018-8-30
Uploading and Loding Picture ...(0/1)
o(^-^)o
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

The manual says that the 4K 28mm Full FOV mode downsamples the image of the 5,7K sensor to UHD.
That seems not to be true. It seems that it skips or bins lines.
The image is very soft and with low resolution in contrast to the Phantom 4 Pro that really seems to oversample by downscaling the image of the probably same 5,7K sensor.

Can this issue be fixed? Is the video SoC an FPGA that can be reprogrammed?
4K HQ mode shows more detail, right, but it should be the opposite (5.7K oversampling should be better) what Phantom 4 Pro or Sony a6500/a7M3 (6K->4K) cameras show impressively.
2018-8-30
Use props
karlo_m
Captain
Flight distance : 349885 ft
Croatia
Offline

Still haven't recieved my Mavic 2 Pro. Could you upload some examples? I don't know if I'll be able to help but I'm really interested in this since mine is supposed to arrive tomorrow.
2018-8-30
Use props
gnirtS
Captain
Flight distance : 2993497 ft
United Kingdom
Offline

Downsampling is always going to be lower quality than native.
That said, without seeing the footage and sharpening settings its impossible to tell.
2018-8-30
Use props
davidmartingraf
Captain
Flight distance : 38060446 ft
  • >>>
Offline

Perhaps DJI will come out with a firmware release to fix the issue. A 5.7 sensor to 3840x2160p seems like there's lots of upside.
2018-8-30
Use props
Blellow
First Officer
Flight distance : 1007808 ft
  • >>>
United States
Offline

Please, post examples.
2018-8-30
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

These are the results and no, downsampling of oversampled sensor readout data improves quality over native sensor readout since at least image degration due to debayering effects become „smaller“, less visible.



EDIT:
Sorry, the 'measurement' here contains a mistake! One ISO testchart should have a size of 2,81% (5,625% in FullHD) per picturewidth in 16:9, not 2,5%!
But, all tests above (M2P 'HQ', M2P FullFOV and P4 Pro 25p) were done at 2,5%, so they are comparable among each other, but not with other tests that have a correct scale.
I´ll do a correct test later.

7E1D6246-0CA3-49D7-976F-5E196C9554AA.jpeg
2018-8-30
Use props
Blellow
First Officer
Flight distance : 1007808 ft
  • >>>
United States
Offline

Yeah, YouTube Billy has noticed it too.  He thought it was acceptable.
I'm on the fence, but really don't want a phantom 5 because it is too big. Thanks for the example.
2018-8-30
Use props
Heavenone
lvl.4
Flight distance : 3563 ft
  • >>>
Germany
Offline

Roland Schulz Posted at 2018-8-30 19:22
These are the results and no, downsampling of oversampled sensor readout data improves quality over native sensor readout since at least image degration due to debayering effects become „smaller“, less visible.

Hi Roland Schulz,

great work! Bad results for M2pro.
Hope they change it like at the beginning mavic pro also had these bad videos..
Would it be possible to do your check also for 2.7k 50p with mavic 2pro?
If it is not too much work it would be great!
(i am also in Germany)
2018-8-30
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

ATTENTION:
Sorry, the 'measurement' above (https://forum.dji.com/forum.php? ... 367&pid=1542249) contains a mistake!
One ISO testchart should have a size of 2,81% (5,625% in FullHD) per picturewidth in 16:9, not 2,5%!
But, all tests above (M2P 'HQ', M2P FullFOV and P4 Pro 25p) were done at 2,5%, so they are comparable among each other, but not with other tests that have a correct scale.
I´ll do a correct test later.
2018-8-30
Use props
Heavenone
lvl.4
Flight distance : 3563 ft
  • >>>
Germany
Offline

Roland Schulz Posted at 2018-8-30 22:01
ATTENTION:
Sorry, the 'measurement' above (https://forum.dji.com/forum.php?mod=redirect&goto=findpost&ptid=164367&pid=1542249) contains a mistake!
One ISO testchart should have a size of 2,81% (5,625% in FullHD) per picturewidth in 16:9, not 2,5%!

Perhaps if you do a new test,(M2PRO) can you do it also in 2.7k 50p
and perhaps difference H264..H265 in 4k?
That would be great.
I will do later a sparpness test with a special test chart.
2018-8-30
Use props
ulfw
Second Officer
Flight distance : 281381 ft
Turkey
Offline

Oh no.  Seems the DJI release way has happened again. This is an exact copy of the Mavic Air Release earlier this year. Camera was mega soft and at the same time mega noisy. Even at iso 100 in broad daylight noisy. Resolution nowhere near 4K.
Had it replaced after three months. New one worked properly.
2018-8-30
Use props
Chasing Light & Shadows
Captain
Flight distance : 221476 ft
United Kingdom
Offline

Hopefully it's software and not hardware causing the softness, although in the whole scheme of things the actual video footage still looks amazing albeit not quite as sharp as the P4P.
2018-8-31
Use props
Thriller
First Officer
Flight distance : 374337 ft
Italy
Offline

I hope DJI fix this issue soon. I can't believe that a new camera with hasselblad design and 1" sensor have this performance.
2018-8-31
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

I also didn´t ever expect this performance where we had much better results with probably the same sensor in an 2 year old DJI design.
2018-8-31
Use props
Chasing Light & Shadows
Captain
Flight distance : 221476 ft
United Kingdom
Offline

I'm holding my full judgement until I've got my mavic and can test personally and knowing that I've got 100% focus and the correct aperture. So far I've not seen any footage that is in focus and soft, only out of focus footage and soft. The above test charts are only from one sample. I've seen loads of sharp footage that is more than acceptable, personally I'm not worried (yet).
2018-8-31
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

Any comments from DJI staff here? Is this "issue" solveable via a firmware/internal 'flashable' FPGA update or is this a hardware design flaw?
I can hardly believe that this is a wanted 'degration'.
2018-8-31
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

Chasing Light & Shadows Posted at 2018-8-31 01:11
I'm holding my full judgement until I've got my mavic and can test personally and knowing that I've got 100% focus and the correct aperture. So far I've not seen any footage that is in focus and soft, only out of focus footage and soft. The above test charts are only from one sample. I've seen loads of sharp footage that is more than acceptable, personally I'm not worried (yet).

The charts are captured „in focus“. I checked that and no, the FullFOV result doesn‘t become any sharper!
2018-8-31
Use props
Andrejs
First Officer
Flight distance : 1730177 ft
Latvia
Offline

I found same issue. I was thinking that it is problem with my settings. But seems that not
2018-8-31
Use props
Chasing Light & Shadows
Captain
Flight distance : 221476 ft
United Kingdom
Offline

Roland Schulz Posted at 2018-8-31 23:04
The charts are captured „in focus“. I checked that and no, the FullFOV result doesn‘t become any sharper!

I wasn't referring to your charts not being in focus, you'd have to be pretty stupid to miss focus a chart. Was referring to having a different sample of the lense and testing it myself, we all know that lenses aren't all made equal and cannot rely on one test sample alone. I'm not saying your results are incorrect, but one test chart isn't enough to draw a conclusion.
2018-9-1
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

Chasing Light & Shadows Posted at 2018-9-1 00:42
I wasn't referring to your charts not being in focus, you'd have to be pretty stupid to miss focus a chart. Was referring to having a different sample of the lense and testing it myself, we all know that lenses aren't all made equal and cannot rely on one test sample alone. I'm not saying your results are incorrect, but one test chart isn't enough to draw a conclusion.

Realize that the „HQ“ crop mode shows a higher resolution than the full sensor mode in my test.
The MTF of the lens is demanded more in the crop mode so it‘s not the lens limiting in the full sensor mode!!
2018-9-1
Use props
Chasing Light & Shadows
Captain
Flight distance : 221476 ft
United Kingdom
Offline

Roland Schulz Posted at 2018-9-1 02:38
Realize that the „HQ“ crop mode shows a higher resolution than the full sensor mode in my test.
The MTF of the lens is demanded more in the crop mode so it‘s not the lens limiting in the full sensor mode!!

I dunno exactly what's causing it then, need to test it myself in real world and see if it's acceptable, from the decent footage I've seen so far it looks like it will be.
2018-9-1
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

It is useable, but I feel a bit ripped of since the 2 year old P4P performed MUCH MUCH better here.
2018-9-1
Use props
Roland Schulz
lvl.4
Flight distance : 2166962 ft
  • >>>
Germany
Offline

No news yet but I did some tests with the 8bit FullFOV standard mode since I guessed the binning could be neccessary because of the 10bit recording/maybe 12bit readout.
Unluckily 8bit recordings also show mediocre resolution in Full FOV, look pixelbinned, unlike the 2 years old P4 pro.

Maybe there is another problem. Sony´s FDR-AX100/AX700/PXW-X70 camcorder, Sony DSC-RX100IV/V and also DJI´s former Phantom 4 Pro all output very good (oversampled, no pixel binned) 4K resolution in full pixel sensor readout mode. We know that 4K recording is somewhat limited on the RX100 series because of temperature rising/overheating.
I never encountered those problems with my AX100/X70 Sony camcorders that also do full pixel readout nor with my Phantom 4 Pro. On the Phantom 4 Pro there may be better air circulation from the rotors to the camera for cooling, the Mavic 2 Pro camera is somewhat covered and so maybe gets less cooling and so maybe overheat in full pixel sensor readout mode.

But DJI, that´s quite simple. Please monitor the sensors temperature in the app and if temperature get´s to high in full pixel readout mode switch to pixel binned mode and show a warning.
We don´t need full pixel readout for plain flight monitoring while not recording, we only need it while recording in Full FOV and I guess not to many people will record video so long that the sensor overheats.
Please implement an option for full pixel readout that was present on the P4 Pro.


By the way the standard color profile is currently hardly useable since it´s much to contrasty, implementation of D-Cinelike would be desirable.
2018-9-2
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules