DJI Mavic Air 2S - First Impressions

Discussion in 'Aerial and Drone' started by Ed_Ingold, Jun 17, 2021.

  1. I don't have an answer, because I don't see it in my Air 2S. I'll look closer at settings. If you find an answer, please share it.

    For what it's worth, the difference between blue and green is extremely dependent on white balance. In part it's because human blue/green sensitivity peaks are very close together. After cataract surgery I found that things I saw as green now are blue.
     
  2. I had a similar experience with cataract surgery because the cataract itself has a yellowish tint. Whites looked blue-ish for a short time.

    I'll have to try with more controlled colors, but I just did some edits that suggest that you're right: it's more likely a WB problem than a green calibration channel.

    Here's an unedited shot fairly late in the summer, so the greens should not be this yellowish:

    [​IMG]

    Now, here's the image with WB adjusted using the tent at the lower right as the target:

    [​IMG]

    The change on the blue-yellow axis is very small--slightly more blue--and is smaller than I noticed in some other photos. But there is also a larger shift from magenta to green.

    Finally, here's the original with the green channel pulled to +40 in Lightroom and no adjustment via the WB control. The change in the greens is more extreme, probably overdone. But the walkway has a clear magenta cast, which is not accurate. So the bottom line seems to be a not terribly good AWB.

    [​IMG]
     
  3. Automatic white balance is greatly affected by dominant colors in the scene. The expanse of green in this image is likely to push the WB toward magenta. Secondly with three primary colors there are only n-1 degrees of freedom. The primary variable for white balance is the red/blue balance, with green mainly a measure of luminance. That's not to say you can't adjust green against its complementary color magenta, but that balance has a lot of overlap with the traditional red/blue balance. Finally, different cameras use different methods to establish the color balance. If it were based on dominant color alone, images would be pretty much useless without a lot of work.

    To be honest, I don't see a striking difference between the images above. By itself, any one might be considered normal and usable. Among a series of similar images, consistency is more important than being "correct."

    The problem I face on an almost daily basis is matching several video cameras, sometimes of different makes and models. Switching between cameras highlights any differences. I start by setting white balance with an 80% white card, which helps minimize these differences. While shooting, I usually have my hand on exposure control too. Using a fixed WB eliminates variations caused by changing scene contents (if the client adds colored lights, I find it best to let it happen). While that addresses the top end of the scale, different cameras have different curves for different colors in the mid range, and professional cameras have many profiles for these curves, depending on the cinematic intent. Hollywood colorists are paid handsomely for this service. To stay within budget and schedule, I have to settle for good enough. Since I have neither schedule nor budget concerns with drone photography, "good enough" is the the only rule.
     
  4. Hi, I'm mainly a still photo guy, going to paper prints, so the requirements won't be quite the same. But the first thing that sticks out to me is a very flat tonal response. In fact, I first thought that your sample shots were taken on an overcast day. But looking closer I see that the trees, etc., have long shadows with hard edges. But the shadows are very weak, almost as though a too-strong fill-flash came from the camera. To my way of thinking these shadows should be getting much darker than they are.

    I didn't examine the images any deeper than this, but this is the first place I would start with these images - to find why the tonal response is so weak. Maybe there is a certain video profile needed, in the same way that ICC profiles are used with still cameras.

    You might try shooting the same scene with both your drone camera and one of your Canons to see the tonal difference. (If in doubt as to the "aim," just use a standard jpg from the Canon.)

    I wouldn't bother looking into color issues too much at this point (although I always am a fan of doing manual white balance when possible). When you start increasing contrast in an RGB system (via "curves," etc.) the color saturation starts climbing with it. Thus my suggestion to deal with the tonal response first. That's how I'd approach it anyway.
     
  5. Bill,

    Thanks for the comments. Some of the drone images I've captured have been somewhat flat. If necessary, I deal with the saturation issue by doing at least some of the contrast and other tonality adjustments with a luminosity blend mode in Photoshop. That works solely on luminosity, so you can adjust contrast as much as you want without having any effect on saturation.

    Saturation and WB are very different things. The WB problem I have noticed is primarily too much yellow mixed with greens. Increasing contrast or saturation wouldn't solve that problem. That's why I originally played with calibration, but Ed was right: it's a WB problem. The images I posted have a pretty mild case of this problem. Some of the other images I've taken have had greens that are quite far off.

    Dan
     
  6. Just for fun I took a closer look at the first image you posted. Here's an odd thing - in RGB the green channel never gets lower than about 35 or 38 in pixel values (but red and blue DO get down close to 0). If you use something like Photoshop "Levels" or "Curves," setting green pixel values from about 35 down to 0, that overly-green foliage becomes more normal.

    From that point the tree shadows still look too weak to me. So I would pull down the overall "Curve" in the low areas. For example, pull a pixel value of about 25 down to about 12 or so. I actually made this be part of an overall "S" Curve which crosses at about 75 (the upper part bulges slightly up). The color then becomes much too strong, so desaturate to taste.

    I think you'd find these changes much improved over the original. But the real issue, aside from the poor tonal response, seems to be the failure of the green channel to go low enough. I have no idea why this would happen; it really shouldn't (this is not a result of white balance settings). Maybe there's some setup information in your camera manual?
     
  7. Thanks for the comments, but the issue I initially pointed out is not one of poor tonal response. The problem is that in some images, greens have too much of a yellow cast, not that they are too green. For example, look at the bottom tree foliage in this one, which is raw read into LR with no adjustments;

    [​IMG]

    That color would have been reasonable in May, but not at the end of July.

    Having now gone over a bunch of drone images taken under different circumstances, I'm pretty much convinced that Ed was right that this is just a poor AWB result. I do have a few images where tweaking the green in the calibration panel also helps. The problem is resolved, I think.
     
    Last edited: Dec 13, 2021
  8. Raw images with a wide dynamic range often look faded compared to their processed versions. This is the unprocessed version of an image I posted above. The advantage of RAW imaging is that you have the latitude to do considerable processing. I used AuroraHD to process this image.

    DJI_0216.jpg
     
  9. Right. The faded appearance was therefore not a concern for me. I'm used to it. My concern was solely color rendition, particularly yellow-green
     
  10. Just as a demo, here's what happens using the sequence I described. (If color is too weak, probably cuz I desaturated too much in the last step.)

    The major change, in my view, was due to the green channel not having any values below about 30 or 35, or so. I used a "Levels" control to drag it down near zero.

    The white balance is essentially unchanged, as evidenced in the rectangular white things on the right. As I said before, I have no idea why the green values didn't go lower - it seems like they should have.

    [​IMG]

    Below is the original, the first one in post 22, above.

    [​IMG]
     
    Last edited: Dec 14, 2021
  11. Now that we're in the three-day interlude between Winter and Summer, I'm finding a few more things to photograph in Illinois. A familiar place is Apple River Canyon SP, in northeast Illinois. It's beautiful in the fall (see above), but not bad this time of year either.

    I decided to add a more sophisticated controller, the new DJI RC Pro. It has a built-in screen, bright enough to use in sunlight, and more ergonomic controls. The original controller needs a smartphone for control and viewing, only visible in the shade. The controller also works with a Mavic 3, and probably more models as time goes by.

    DJI_0294_AuroraHDR2019-edit.jpg
     
  12. One of my friends bought a rack that allows him to use an iPad rather than a phone with the standard controller and says it's a big improvement. I haven't tried it yet.
     
  13. A farm in western Massachusetts this week:

    [​IMG]
     
    Ed_Ingold likes this.

Share This Page

1111