DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Auto exposure gives better highlights and shadow details in high ISO than manual at lower ISO

fnf

Active Member
Joined
Apr 21, 2017
Messages
39
Reactions
33
Age
37
In recent history, DJI's drones have always been favouring high ISO and high shutter speed rather than sticking to as low ISO as possible and lowering the shutter speed to compensate. I'd always found that puzzling.

The recommendation that most people give is to disregard the drone's poor auto-exposure algorithm and go with manual exposure instead, such as in this thread Why is it bumping base ISO100 even at shutter speed of 1/3000? .

As I found out, there is a very good reason as to why the drone does that: highlight recovery. In essence, DJI's exposure strategy usually tries to preserve at least 2-3 stops of highlights then bumps up the ISO to render a more balanced exposure. The result is a slightly noisier image but with less highlights being blown out. As a result, videos taken at higher ISO can be of better quality than lower ISO.

Here are screen grabs from a video taken with Auto ISO/SS at ISO 1600 and 1/500s (Pro mode):
Screenshot 2025-03-10 161947.jpg

and here's the exact same scene with manual ISO 240 and 1/60s shutter speed:
Screenshot 2025-03-10 162512.jpg

Notice how both highlight and shadow details are gone from the manual exposure one.

The loss of highlight details were due to the reason above, but the loss of shadow details were likely caused by the noise reduction algorithm (I use Sharpness -2, Noise Reduction -1) and the extra noise helped avoiding it.

I do not want to use Noise Reduction -2 since it causes unnecessary loss of details in general due to bitrate budget. NR -1 gives the best compromise for me in terms of detail retention.

So, as surprising as it is, ISO 1600's picture quality was a lot better than ISO 240. Many recent DJI sensors also have dual native ISO, so high ISO could still look very clean and is an extra reason why we should use Auto exposure with EV to compensate.

I have only tested the Air 3S, Mini 4 Pro, Avata 2 and Neo thus far, but all of them provide better picture quality with Auto ISO/Shutter speed enabled. The best exposure strategy would be ETTR with Auto ISO/SS, rather than manual exposure like with other cameras.
 
I think you have to figure out if this is an ISO thing or an auto exposure thing before we jump to conclusions on how to solve it. The test for that would be to take two examples, one set with auto exposure and then the second set to the same settings but manually and see if you get the same differences.

-If same difference, then auto exposure is applying some other processing we don't know about.

-If not, then it has something to do with the ISO value possibly the a dual gain or even more possibly it has something to do with the quad bayer sensor which is always the wild card on these drones.
 
  • Like
Reactions: Dave Maine
I think you have to figure out if this is an ISO thing or an auto exposure thing before we jump to conclusions on how to solve it. The test for that would be to take two examples, one set with auto exposure and then the second set to the same settings but manually and see if you get the same differences.

-If same difference, then auto exposure is applying some other processing we don't know about.

-If not, then it has something to do with the ISO value possibly the a dual gain or even more possibly it has something to do with the quad bayer sensor which is always the wild card on these drones.
Yes, Auto ISO mode applies some HDR-style post-processing to the video feed which isn't done if the same ISO/shutter speed are set manually. It's easy to see for yourself by toggling it. You should observe an immediate difference in the highlights rendered. This is a known fact about the Neo (Avata 2/Mini 4 Pro do behave slightly differently but Auto exposure mode still give better results in different ways).

This post-processing should produce higher-quality video since the drone would be working with raw data. Doing so in post won't be as good due to compression artifacts. Noise will be present regardless if shadows are lifted in post, but if we shoot in low ISO to begin with to preserve highlights, there might be no shadow details to recover due to compression.

My post was focused more on seeing whether it's worth accepting the higher noise in the shadows in exchange for better highlight details. Interestingly enough, the shadows are better in this case too so it's a win all around.
 
Last edited:
To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.
Any particular point in the video you'd want to refer to?.

Edit: found it (~7:20), this aligns with my observation, thanks.

Many modern cameras (particularly phones) employ the same trick. They seem to be based on/inspired by the Dual Conversion Gain technology from OmniVision.
 
Last edited:

DJI Drone Deals

New Threads

Forum statistics

Threads
136,080
Messages
1,613,260
Members
164,651
Latest member
cwardster
Want to Remove this Ad? Simply login or create a free account