Advice for a novice drone operator looking for best stills image quality.

[noosaguy]

I agree, and said as much in my first post.

I didn't get the Mavic because I thought it was better than my D850. I got it because I thought I could get it to places where I can't get my D850, and because I thought that, with a little digital work, I could get results that were good enough for my liking. I carry my RX100m3 in my car instead of my larger gear because it's small, expendable, and is capable of decent results - certainly better than my cell phone. The best camera is the one you have with you.

That said, having some actual quantifiable data on what the M2P can do may be useful, at least for some people. Lots of people in other forums are arguing about what the bit-depth of the sensor is. That's a pretty easy question to answer: it looks almost certain that it's 12-bits. Lots of people are asking about the low light performance and the shadow detail compared to other 1" sensors. Again, I see a lot of people arguing about it, but I don't see a lot of people taking an actual look at side-by-side results coming out of cameras in semi-controlled conditions.

How might this be helpful? Well, if it looks like the M2P has somewhere between 2-3 stops less DR than what I'm used to shooting, that means I might want to go for a 5 shot AEB bracket instead of just three. In fact, I might want to do a manual bracket with 5 shots at 1EV intervals instead of using the automated function to get that extra 1.5 stops of DR.

My intent isn't to pixel peep - my intent is to know as much about the camera on the M2P as I can so I can use it as effectively as possible. I know it's not going to be as good as what I'm used to shooting with. What I want to know is what the relative performance is going to be so I can compensate for that as best as possible.

And, if we're honest, I'm a professor of electrical engineering... I kind of can't help it.

Cheers.

Here are the 5 AEB shots I used to process in LRCC HDR. Not 5 stops but it seemed to do the trick. Certainly the lighting was pretty good though so under more demanding conditions manual bracketing might be better as you say.

 

[Bussty]

If you would like the original picture in JPEG so you can print it and see what the Mavic is capable of I'd be happy to email it to you so you can play with it..

Hey Neggy thanks for the reply and really glad to hear you can enlarge so much. I have a couple of RAW files now but thanks anyway for the offer. Seeing your images it really is astounding just how many uses these drones can have!

Cheers

 

[Bussty]

Dynamic range will be objectively worse than your M43 gear, to say nothing of comparing it to a newer full frame (D850, A7Rm3). I've not done a super in-depth analysis, but my guess is the Mavic has about 2-3 stops less dynamic range than the OM-D E-M1 / GH4 generation of M43 gear, based on a quick test yesterday.

The files coming off the sensor have a 16-bit output, but I'm fairly certain we're dealing with a 12-ish-bit signal path that's highly biased (black level appears to be 4,096). After accounting for the black level, you've got two bits (2 and 3) totally unused, and only minor activity in bits 0 and 1. I'll probably try to do some more analysis over the next few weeks, but that's my initial read.

From a more subjective "looking at the pictures" perspective, when looking at "the same" scene quickly captured with the M2P, a GH4 and a D850, it's clear that the Mavic is the worst performer of the three. Obviously the D850 was going to win that contest, but I expected the M2P to put up a better showing compared to the GH4. The M2P has significantly more noise in the shadows, even though the GH4 image was pushed about 1.5 EV more. For reference, I've included a shot in the shadows for the MP2 (pushed 2.4EV), the GH4 (pushed 4.2EV) and D850 (pushed 3.3EV) below. All of these are using the base ISO for the sensor in question (100, 200, 64, respectively). Again, this was a (very) quick test, but it should give you some idea of the relative noise performance in the shadows. I may try to put together something more comprehensive in the coming weeks.

I've not had a chance to get the M2P out to do serious work, so it's hard for me to say how right now how it will fare in more challenging conditions, but my initial read is that it's not going to be as forgiving as a more modern (edit: larger) sensor in terms of exposure. That said, with proper ETTR technique, you can get some decent results. The image below is a stitch of 6 shots from a sunset that fizzled a bit on Friday night. My guess is that more complicated scenes will need AEB with some HDR processing on the back end to get similar DR to what is possible with a modern full-frame, but in theory you ought to be able to get 3 extra stops with DJI's 5-shot AEB (a 0.7EV bracket).

The bigger thing, of course, is that, as others have said, you can get the M2P places where it's just not possible to get a D850, at least without a private helicopter and a good bit of cash. Certainly the sensor performance is a little disappointing for those of us who have cameras that are much more capable, but the strength of the Mavic is where it can go, not its sensor.

Thanks so much for going to that trouble. I was actually quite surprised how much noise was in the m43 GH4. Having seen your results and others I'm actually quite happy. If I encounter noise I can stack and use a mean command to smooth out and if I need more resolution I guess I can stitch a group of images. Would be so cool if the M2P had a mode that physically moved the drone closer then did a 9x9 grid stitch to then effectively give you the same field of view but with a super high resolution image, a bit like the Mavic 2 Zoom does but physically moving in rather than zooming (cos it can't). I think if the developers really wanted to they could do that through firmware and it certainly would level any concerns about not being as high res as a m43 or full frame. What more would I want? (oh just through in AEB on the 9x9 grid stitch and woaaa!)

 

[jwischka]

Thanks so much for going to that trouble. I was actually quite surprised how much noise was in the m43 GH4. Having seen your results and others I'm actually quite happy. If I encounter noise I can stack and use a mean command to smooth out and if I need more resolution I guess I can stitch a group of images. Would be so cool if the M2P had a mode that physically moved the drone closer then did a 9x9 grid stitch to then effectively give you the same field of view but with a super high resolution image, a bit like the Mavic 2 Zoom does but physically moving in rather than zooming (cos it can't). I think if the developers really wanted to they could do that through firmware and it certainly would level any concerns about not being as high res as a m43 or full frame. What more would I want? (oh just through in AEB on the 9x9 grid stitch and woaaa!)

Keep in mind the GH4 shot is pushed over 4EV - the original shot was trying to capture both the entire range of the very brightly lit outside with the dimly lit interior, so some noise in the shadows was expected.

 

[STKNRUD]

I love this thread. Good info from experienced photographers using a Mavic. That alone tells me it takes some pretty good photos. But I laugh when the M2 is compared to a full frame Nikon and lenses that can’t fly and hurts your back to carry very far with a pro tripod.

 

[kilomikebravo]

That alone tells me it takes some pretty good photos.

Indeed it does, despite the complaining here in the forums. I have a 7D with a 15-85 and one expects good performance but I have to say, I was totally blown away with the image quality of stills from the M2P.

 

[mbre]

The image below isn't from my M2P it was taken a couple of months ago with my 80D out of a chartered helicopter. The point is though, the charter cost me $700 and I reckon I could have achieved that easily with the M2P for nothing.

Great shot! Where is it?

 

[noosaguy]

Great shot! Where is it?

Hey MBR thanks for your comment. I'm surprised you don't recognise it, it's Uluru in outback NT Australia. It used to be called Ayer's Rock and it is an iconic tourist attraction in Australia. To give it scale, it is 350m high and has a 9.5km circumference.

 

[jwischka]

I love this thread. Good info from experienced photographers using a Mavic. That alone tells me it takes some pretty good photos. But I laugh when the M2 is compared to a full frame Nikon and lenses that can’t fly and hurts your back to carry very far with a pro tripod.

Some of you guys (including dtp who has since deleted their post) need to realize that comparison doesn't necessarily mean criticism. As I mentioned in a reply to noosaguy, the point of comparisons - at least as far as I'm concerned - is to figure out how best to use the M2P, not to put it down. Too many people seem to be overly defensive about their purchase, and hence attacking anyone who doesn't give the M2P 5 stars for its image quality. I'm more interested in learning how to use it to the best of its abilities.

This weekend I was able to get it out and shoot in some real conditions. My initial impressions are favorable. There's no question the sensor is capable of some decent images. The attached images are first-cuts of processing I did on the plane on the way back, but give some idea of what the sensor is capable of. My initial impression in real world conditions is that the M2P may be more forgiving with highlights than shadows - I don't have any hard data on that at the moment, so I put it out there somewhat cautiously. I was impressed, though, by how much I was able to pull the highlights back, generally, without losing detail.

 

[noosaguy]

Some of you guys (including dtp who has since deleted their post) need to realize that comparison doesn't necessarily mean criticism. As I mentioned in a reply to noosaguy, the point of comparisons - at least as far as I'm concerned - is to figure out how best to use the M2P, not to put it down. Too many people seem to be overly defensive about their purchase, and hence attacking anyone who doesn't give the M2P 5 stars for its image quality. I'm more interested in learning how to use it to the best of its abilities.

This weekend I was able to get it out and shoot in some real conditions. My initial impressions are favorable. There's no question the sensor is capable of some decent images. The attached images are first-cuts of processing I did on the plane on the way back, but give some idea of what the sensor is capable of. My initial impression in real world conditions is that the M2P may be more forgiving with highlights than shadows - I don't have any hard data on that at the moment, so I put it out there somewhat cautiously. I was impressed, though, by how much I was able to pull the highlights back, generally, without losing detail.

I appreciate your view, however in my opinion, those images are not able to be judged in regards to performance. IMO they are way over saturated with way too much contrast. I know some people like their images that way and they have every right to, but to present them as being an example of what the m2P is capable of producing image wise isn't going to cut it (IMO).

 

[Meta4]

The attached images are first-cuts of processing I did on the plane on the way back, but give some idea of what the sensor is capable of.

Left in the oven a little too long?
The overcooking has hidden what the sensor might have delivered.

 

[jwischka]

I appreciate your view, however in my opinion, those images are not able to be judged in regards to performance. IMO they are way over saturated with way too much contrast. I know some people like their images that way and they have every right to, but to present them as being an example of what the m2P is capable of producing image wise isn't going to cut it (IMO).

Left in the oven a little too long?
The overcooking has hidden what the sensor might have delivered.

Actually you're looking at (only) local exposure adjustments (exposure, highlights and shadows) and a global contrast on the image to +75, blacks to -10, clarity to +15 and dehaze to +5 on the sunset. The saturation and vibrance sliders weren't touched. Similar settings on the landscape, though saturation is +5 in LR.

SOOC from a single layer of the sunset, for reference. Ultimately the HDR didn't provide a whole lot of benefit (the two are virtually indistinguishable with similar settings). And with contrast pulled back to 25, if you'd prefer to see it a bit more muted.

Also SOOC and identical to above with contrast pulled to 25, saturation zeroed for the landscape.

This is basically just exposing to the right and evening the dynamic range across the image.

 

[noosaguy]

Actually you're looking at (only) local exposure adjustments (exposure, highlights and shadows) and a global contrast on the image to +75, blacks to -10, clarity to +15 and dehaze to +5 on the sunset. The saturation and vibrance sliders weren't touched. Similar settings on the landscape, though saturation is +5 in LR. SOOC from a single layer of the sunset, for reference. Ultimately the HDR didn't provide a whole lot of benefit (the two are virtually indistinguishable with similar settings). And with contrast pulled back to 25, if you'd prefer to see it a bit more muted. But seriously, there's really no color pushing going on there, outside of contrast adjustment. Just exposing to the right and evening out the levels across the image.

I have nothing to add.

 

[jwischka]

I have nothing to add.

I'd be happy to send you the DNG, if you'd like to process it and post your results. (No snark here - I'm genuinely interested in what you'd do with it).

 

[noosaguy]

I'd be happy to send you the DNG, if you'd like to process it and post your results. (No snark here - I'm genuinely interested in what you'd do with it).

Look I'm not having a go at you in any way. I don't have the right to pull apart another's images unless I'm asked specifically to pass judgement. Whilst I don't particularly like images processed the way you have, there will be lots of people that will think they're great. Everyone has an opinion and a right to do it their way.

Notwithstanding that, I don't believe you can't post over-processed images and ask people to make a value judgement on a sensors (cameras) capability. That's why i commented in the first place... not to criticise.

You can send the original DNG for me to process if you like but it seems pointless because personal opinions vary. Some will like the way you have done it, some will like the way I would do it. Basically it doesn't mean a thing.

 

[jwischka]

Look I'm not having a go at you in any way. I don't have the right to pull apart another's images unless I'm asked specifically to pass judgement. Whilst I don't particularly like images processed the way you have, there will be lots of people that will think they're great. Everyone has an opinion and a right to do it their way.

Notwithstanding that, I don't believe you can't post over-processed images and ask people to make a value judgement on a sensors (cameras) capability. That's why i commented in the first place... not to criticise.

You can send the original DNG for me to process if you like but it seems pointless because personal opinions vary. Some will like the way you have done it, some will like the way I would do it. Basically it doesn't mean a thing.

I don't disagree regarding personal styles. My intent was to say, "Here's a scene with a lot of dynamic range and natural contrast taken with the M2P. If you're taking pictures of scenes like this, this is the range of detail in the various parts of the image you can get."

Obviously processing is a personal thing. As you say, some will like mine, some will like yours. But if your contention is that you can't just post an overprocessed image and make a value judgment, I'd be happy for you to post your own version so people can have a broader range of what is possible.

Dropbox - DJI_0807-Pano.dng

 

[noosaguy]

I was unable to download a full size version but I pulled your 2500px image and wound things back a tad. As I said, it's a pointless exercise because of personal preferences. I like mine, you like yours. To be fair and make value judgement on what a sensor is capable of, the original RAW image is the best option. I bit hard in this case because it's a multi image exposure. If your other file was the original, it was underexposed and it is amazing what could be pulled from the shadows.

Nice location, where is it?

 

[GeraldV]

Hi there

I don't yet own a drone but are looking at getting one. My sole driver is stills landscape photography. I currently use an Olympus EM-1 Mark 2 and some pretty sharp lenses and am really looking to get close to that quality but from the air.

I was really excited when I heard about the Mavic 2 Pro and was pretty much ready to fork out the New Zealand $2600+ as soon as it was announced.

However I waited and have now seen RAW files plus the massive amount of reviews on the Mavic 2 Pro and no longer very keen. I am worried if I buy this drone the quality difference compared with my Digital Camera will be too much and I will be disappointed.

One thing that has really grabbed my attention is the leak of the possible Phantom 5 and the photo of the interchangeable lenses. As you probably all know it's pretty hard to make a bad standard prime lens and my thinking is if the leak is real that we could be looking at a paradigmal change in image quality and sharpness if just only from a lens optical point of view.

I know we are just guessing but my real question to the drone community here is do you think my logic is valid or flawed?

Many thanks for your comments

Cheers

Andrew

I have a mavic air but like you am more interested in photography than video and so uograde to the m2p. The camera quality is stunning for a drone. With the aparture you have a lot of control over the image. The drone is extremely stable and being a drone you can get to angle ww that otherwise would be impossible. Its never going to rival a dlsr bit its better than a high end phone camera. These two pics were taken on the maiden flight. I can send you the raw files if you want to check them out.

 

[jwischka]

I was unable to download a full size version but I pulled your 2500px image and wound things back a tad. As I said, it's a pointless exercise because of personal preferences. I like mine, you like yours. To be fair and make value judgement on what a sensor is capable of, the original RAW image is the best option. I bit hard in this case because it's a multi image exposure. If your other file was the original, it was underexposed and it is amazing what could be pulled from the shadows.

Nice location, where is it?

Park City, Utah.

The DNG is a 6-image stitched panorama, which should function for all intents and purposes like a single exposure at this point. The file is not, in fact, underexposed (at least if you're trying to stick with a single exposure, and not do any HDR) - any higher exposure would have lost detail in the detail in the clouds (specifically the saddle area in the mountain range just to the right of center), which would have been unrecoverable. Expose to the right. This was my point in posting that image, in fact: you've got a very wide dynamic range scene where the highlights are just at the point of clipping. How much detail is still left in the shadows? It turns out a reasonable amount, more than I would have expected based on my initial indoor tests. The highlight detail and recovery is likewise better than I would have expected based on the initial tests. This is what the sensor is capable of.

If you copy the DNG link and replace the "0" with a "1" at the end, it will automatically download.

 

[noosaguy]

Hey jwischka, I reckon we've done this to death. The image is a nice one and a great location. I'd be very happy to call in my own. We may have to agree to disagree on the way we'd finally present it. Keep getting out there i'd like to see more of your work.