Hi there

This is a copy of a message I've sent to DJI support but I thought I'd share it here too – it potentially has huge significance for those of us who want to use the X7 for aerial photography.

I've recently upgraded from the X5S to the X7 and, like many photographers who were looking forward to beautifully crisp 24 megapixel RAW photographs from the X7, I've been left bitterly disappointed.

Compared to the X5S which produced very sharp RAW files, the RAW files from the X7 are so soft that by the time they've been sharpened to a comparable level, all the detail has been crushed into a blurry pulp.

I initially attributed this to the possible use of an anti-aliasing filter on the X7 sensor as it was presumably intended primarily as an aerial filming camera.

However, I then came across this interesting post on RCGroups which gives me hope:

https://www.rcgroups.com/forums/ ... 457&postcount=14078

Other than switching from video to stills mode, the images linked to above were taken at virtually the same time and with virtually the same settings (albeit at f/4.5 for the video and f/2.8 for the still), and yet the video screenshot contains so much more detail – the level of detail I've come to expect from DJI's high-end aerial gimbals.

I've attached a direct comparison of 100% crops from the 6K video and a RAW file so you can see the enormous difference in detail – I can't deliver images with so little detail as the RAW image on the right to my paying clients!

Please can you let me know when you'll be able to address this issue in a firmware update as it's obviously nothing to do with the sensor or lens, but how the sensor data is processed?

This is really crippling my use of the X7 for stills and I REALLY hope it can be addressed soon.

Many thanks for your help and I look forward to hearing from you,

David

Some have noticed a difference between the image on their tablet's screen when using the focus peaking (red outlines) between the video and still mode on distant objects.  It seems to shift for some reason.  I've noticed it at times.  Don't know why it happens, but I doubt it is AA filter related as I've seen it happen in the X5S as well.

fansa84fe8a4 Posted at 2018-10-2 05:33
Some have noticed a difference between the image on their tablet's screen when using the focus peaking (red outlines) between the video and still mode on distant objects.  It seems to shift for some reason.  I've noticed it at times.  Don't know why it happens, but I doubt it is AA filter related as I've seen it happen in the X5S as well.

Thanks for the reply.

I've experimented with both tapping to autofocus and 'calibrating' the lens by tapping on a distant object and the result is the same.

Never checked the focus peaking though ... maybe this will throw some light onto the issue ... perhaps the lens is simply being misfocused in stills mode? That's certainly what the output looks like ... will investigate when I have a moment.

I've logged this as a support request and I will be chasing them weekly for updates.

I've received these replies so far:

"We understand your Frustration regarding this matter, we will make sure to forward this feedback to our Developers so that may be on the next firmware update I will be fixed and we will escalate this concern to our (FAE) Field Application Engineers. They will conduct an investigation about this issue and will handle your case. May I request to provide us the information below so we can forward it to our Field Application Engineers and process for this concern.

Firmware version of your Inspire 2:
DJI Go 4 version:"


and then later:


"We will use this information and forward it to our Field Application Engineers to address this concern. You will be contacted afterward when we get an answer from our Field Application Engineers that we will relay to you.

Hoping for your patience."


Will see ...

Has anyone else been very disappointed by the X7 image quality in RAW photo mode? If so, please post here to throw more weight behind my request to DJI to get this MAJOR issue fixed ASAP!

Many thanks

David

Hey David-

Generally bad form to repost other folks stuff (mine) on other forums...best way to handle it is to simply link to the original post on RCgroups make your comments and leave it at that. I intended that post to be seen there, not here (or I would have posted it here). Also...as I stated over there, the information is anecdotal at best. I'm not sure the conclusions would be born out by rigorous testing. In other situations I've not seen the same level of softness as this early test.
Also - The area you've zoomed in on is not where the camera was focused. (the focus point is the sandstone cliff in the foreground). If you notice, the still shot was taken at f2.8, and the video shot was at f4.5. The Still shot is also at iso 200 vs 100. Note I mentioned in my original post that due to this, it wasn't conclusive...but you're choosing a background element of a shot at f2.8 isn't really a good comparison, and is in fact, pretty misleading. Today with the x7, I'd be testing for detail with the lens set at 5.6-8, as experience has shown that my results have been the sharpest

It appears you have an x7 to test this. Have you done some side by sides of raw stills and raw video that you'd like to share? It would great to have you contribute to the data pool surrounding this issue, as it seems important to you. Also...If you wouldn't mind removing the link to my dropbox, I'd appreciate it. Thanks.

Barry

Hi Barry,
Thanks for your reply and I'm sorry for any bad feelings I've caused by reposting your content on here ... but it was the first glimmer of hope I've seen that the X7 sensor can actually capture sharp content when zoomed into pixel level – regardless of what f-stop was used.

I've been bitterly disappointed by the apparent sharpness of the stills I've taken so far with my X7 – they're so soft that to get useable results I have to apply so much sharpening that the resulting image looks like a watercolour painting – the attached image shows a central 100% crop from an aerial photo taken with the 35mm lens at f/8 before (left) and after (right) sharpening in Lightroom. That was using autofocus by tapping on the building.

I did some tests at different apertures with the 16mm and found that between f/5.6 and f/8 seemed to be the sweet spot, but even then, stills – whether autofocused or manual focused having calibrated the lens to infinity – are still woefully soft.

This level of softness, taken at the sweet spots of the lenses I've used so far (not tried the 50mm yet but see no reason why it would be any better), seems very comparable with your sample image which I reposted above – and yet the crop from the video screenshot you posted shows far more detail – comparable to the level of sharpness I get from RAW stills taken with my X5S.

Hence my deduction that this isn't simply an issue with my copy of the X7 or the lenses, but a more widespread issue, possibly related to firmware and the way the sensor data is processed for stills.

Whilst I do have an X7, I don't have SSD drives or the RAW / ProRes licences otherwise I would have been knee-deep in making comparisons and filling this thread with them! Your comparison was the only one I've been able to find anywhere on the internet which directly compares the same scene taken as both a RAW DNG still and a screenshot from a RAW video clip – which is why I was so excited when I found it!

And yes this matter is important to me as I use my Inspire 2 as the primary aerial platform for my business, and having spent over £6,000 on a new camera and 3 lenses which promised a 33% increase in image resolution (at the 3:2 aspect ratio) when compared to the X5S, you can imagine my frustration on discovering the IQ has actually gone down, significantly!

Based on your sample and my experience of taking RAW stills, i think it's a fair conclusion that there is an issue with the way the Inspire 2 processes the sensor data when in RAW mode, which is why I started this thread (and contacted DJI directly) to bring this significant flaw to their attention.

I've removed the link as requested above and removed the text referring to your DropBox.

Apologies again if I offended or upset you by sharing this content on here – it couldn't have been further from my intentions.

All the best and thanks again for sharing your findings over on RCGroups – here's hoping it's the catalyst for DJI to fix this issue for all those who use the X7 primarily for taking aerial photographs. The dynamic range, low noise and scope for post-production of the X7's RAW files is certainly impressive, and if I can get RAW files with the level of detail contained within your video screenshot, I will feel the upgrade was justified. At the moment I simply feel cheated by DJI that they've released a bug-laden product into the market place – particularly when combined with the equally rubbish moire banding problem reported here which has already ruined one of my paid shoots.

David

At the bidding of another user at RCGroups, I did a pixel level detail comparison between the x7 and My 5dsr. The results were to put it mildly...confusing.

The 5dsr had much higher noise levels, and much higher default sharpening  (thus aliasing and artifacts) but slightly less high contrast detail.
the x7 had much lower noise levels (almost none), and had no apparent sharpening applied (no aliasing or artifacts) and had equal or better high contrast detail.

The confusing part came when I looked at some of the areas with low contrast detail. The 5dsr seemed to overemphasize this detail (due to the sharpening) and the x7 appeared to be smearing that same detail. In particular, the green foliage seemed to have a lot of detail smearing on the X7 and too much texture on the 5dsr. To me...all of this adds up to aggressive noise reduction on the x7, and maybe not enough on the 5dsr (this camera has been criticized for it's noise levels, although I've always felt that Canon has a more hands off approach to NR than some other manufacturers.)

The 5dsr has a canceled aliasing filter, so we'd expect higher sharpness and some aliasing artifacts. The x7 has an AA filter that is tuned for cinema imaging, so we'd expect a softer image, less artifacts and perhaps the smearing of some low contrast detail. Based on the above --this is generally what we are seeing, although aggressive noise reduction may be making the difference more extreme.

I will say that viewed at 100%, the still image from my x7 does NOT look soft. The detail smearing I mentioned above is only visible at 200-400%.

Here's a night and day comparison of why I'm so disappointed with the X7 when compared to the X5S for stills.
The X7 RAW image basically has a third of the clarity of the X5S images (I tried scaling the X7 down by 33% and then 300% and the level of detail is then equivalent.

The X7 image was taken at f/8 which in my tests appears to be the sharpest aperture for the lens, and by tapping to auto-focus on the building itself – so the image isn’t simply out of focus.


Awaiting a response from DJI ...  

David

I would love the x7 to be sharper its a very expensive camera to be as soft as it is. the 35mm lens here in Australia is $2200 and you can pick up the 45mm Olympus for $450 and the Oly is a 1.8, the DJI is closer to 3.2..
I would say all of my shots are similar to yours in sharpness, I had just accepted the way it was until I had seen the 100% crop from your video, looks very sharp, I would also like to hear fro DJI on this, if they could fix the sharpness for stills whilst fixing the banding issue that would great..

Yes that 6K video screenshot has given me huge hope that the soft raw photographs are just down to a processing problem and not a sensor / lens issue!

I'd love to see some more examples of screenshots from raw 6K video as that seems to give us the best interpretation of the real resolving power of the X7 and its suite of lenses ...

In the mean time I'll be pestering DJI once a week until I get an affirmative answer ...

David

In previous side by sides, with equivalent lenses. ie -- X7 35mm to x5s 25mm -- I've seen that the x7 has the same or slightly better detail in the center, and significantly better detail in margins and corners compared to the X5s. Not saying your test is inaccurate or skewed, but side by sides are the best...take the same image at the same (relative) time at  equivalent focal lengths at optimal apertures. Then  we can compare apples to oranges, not zebras. I find setting the drone on an elevated platform (uhh...table), with a res chart included in the frame, to be helpful. (My tests with the x5s always showed a certain amount of aliasing, which can be confused with detail, and a chart will usually disabuse that notion, and shooting with the craft in the air is a test of the entire system, not the lens/sensor/processing).

We might also want to consider whether your camera/lenses have an alignment issue. It may be worth opening a support case.

Barry Goyette Posted at 2018-10-8 09:24
In previous side by sides, with equivalent lenses. ie -- X7 35mm to x5s 25mm -- I've seen that the x7 has the same or slightly better detail in the center, and significantly better detail in margins and corners compared to the X5s. Not saying your test is inaccurate or skewed, but side by sides are the best...take the same image at the same (relative) time at  equivalent focal lengths at optimal apertures. Then  we can compare apples to oranges, not zebras. I find setting the drone on an elevated platform (uhh...table), with a res chart included in the frame, to be helpful. (My tests with the x5s always showed a certain amount of aliasing, which can be confused with detail, and a chart will usually disabuse that notion, and shooting with the craft in the air is a test of the entire system, not the lens/sensor/processing).

We might also want to consider whether your camera/lenses have an alignment issue. It may be worth opening a support case.

For what it's worth. I went back and examined a group of images for which I have both raw video and stills (a vineyard with lots of high frequency detail to examine). I see no significant difference in detail/sharpness between files shot as video versus still on the x7. (there is a marked difference in color, but that's another issue). There is one curious still file that seems softer than the others, probably a focus issue...but otherwise I don't see anything untoward going on.

I'm Curious David if you've resolved your issue, or been able to do a side by side with your x5s and x7...

It's on my to-do list!

Meanwhile this seems to be the best I can get out of my X7 / 16mm at the moment ... taken yesterday.

http://horizonimagingportal.co.u ... sample_DJI_0525.jpg

and a sample from my X5S / 12mm taken a few months ago:

http://horizonimagingportal.co.u ... sample_DJI_0164.jpg

Both have sharpening and noise reduction set to 0 in Lightroom.

This is actually the best I've seen from the X7 so far, but it still doesn't feel quite as sharp as the X5S as you can see from the 100% crops:

http://horizonimagingportal.co.u ... ture_comparison.jpg

You have to push the X7 image harder to get the same level of sharpness and as a result the fine detail in things like foliage get lost. Whereas the X5S image is full of detail straight out of the camera, and with some mild sharpening it becomes razor sharp!

I know I'm getting into pixel peeping territory here but for the price difference between these two setups I think I'm allowed to pixel peep

On the shoot yesterday I was tapping to autofocus with every image rather than just relying on a single calibration and then setting the lens to infinity focus ... although that's what I did a few posts above with the 35mm lens where that brickwork was so blurry ...

Will try to do some test shots on the ground using the same target on both setups.

Thanks for your input

David

Barry Goyette Posted at 2018-10-10 16:13
For what it's worth. I went back and examined a group of images for which I have both raw video and stills (a vineyard with lots of high frequency detail to examine). I see no significant difference in detail/sharpness between files shot as video versus still on the x7. (there is a marked difference in color, but that's another issue). There is one curious still file that seems softer than the others, probably a focus issue...but otherwise I don't see anything untoward going on.

I'm Curious David if you've resolved your issue, or been able to do a side by side with your x5s and x7...

So do you think that original shot which you posted on RCGroups was a one-off, possibly caused by a mis-focus? The difference seemed startling in that example!

Thanks for your help

David

Hogster Posted at 2018-10-11 00:58
So do you think that original shot which you posted on RCGroups was a one-off, possibly caused by a mis-focus? The difference seemed startling in that example!

Thanks for your help

So as for your samples...I see what you're seeing , but it's not definitive -- as things like the distance and types of subject matter, lighting etc are all different. What I do see is that the x5s has a fair bit more noise, and that, in combination with higher aliasing, could be introducing the "appearance" of detail in the shot...sure I see that things like the bricks on the driveway are rendered, and that's typical of a sensor with a weak or no AA filter.  When I sharpen the X7 image with a .3-.5 pixel radius, I get a result that looks very similar to what the x5s looks like...it's a little more aliased and noisy. and at 100%.. you'd be hard pressed to pick out which image was which. At 200-400% I'm seeing problem with both photos...but that's what pixel peeping is about...finding problems. 100% generally is the only one that matters. I'll bet if you shot the same image, with both cameras, sharpened the x7 a "little" and printed both at 16x20, and showed to a client, they couldn't tell the difference...although, they might notice the better color range, and dynamic range of the x7 and if they were really savvy, the lower noise.

As for my experience. I think I already mentioned that the still shot you picked and then cropped to an out of focus area, was focused to an area of the foreground, and that the still shot was at f2.8 versus 4.5 for the video frame. (2.8 is definitely a tad soft on all the DL lenses).  With the vineyard shots I mentioned, the f-stop was the same, and the distances of the shots I looked at were the same, and I didn't see any significant difference in the level of detail..if anything the video shots seemed to have a little more noise reduction.

For perspective, my other cameras are a hasselblad H5d-40 and a Canon 5dsr. Both cameras without AA filters. I'm at least as sensitive as you about the softness introduced by the AA filter and noise reduction on DJI's cameras. I just think the X7 is operating about as designed (as a Cinema camera with a well tuned OLPF), and while your expectations may be more, it's not a defective system. It's just not one designed for higher res stills than the x5s (the vertical resolution of the two cameras is nearly identical, so expecting "more detail" because of the increased cost isn't a healthy expectation).

The question comes then, if it is only "as good" (or not quite as good :-) at rendering aliased detail as the x5s, and that's what you want in a picture...(I'm surprised in all the pixel peeping you haven't noticed the noise of the x5s, and aren't commenting on how noise free the x7 is)...the question is whether the x7 is a better camera for you. Well it does have more horizontal coverage, and the lenses generally are better in the corners...It has about a stop of additional DR, and a significantly better color depth...but if those things aren't as important as pixel level sharpness, and most importantly, you're not using the camera for what it's designed for...video/cinema/le film...then it's probably not the camera for you.

Hi, I'm Italian. I have the x7 with all lens. I too find your problem the same. I think that the X7 has problems with focus to infinity. Auto focus is often wrong.
in fact I always focus manually.

I'm experiencing the same issues with focus.  Sharp images are intermittent at best.  I have both the 16mm and 24mm lenses, the 24mm I thought was a little better but I shot with it two days ago and only 1/3 were acceptable.  I've had other issues that have been resolved (poor communication between lens and camera) with DJI replacing the lens.  The other day I had to stop using my Inspire2 and change to the Mavic 2 Pro which seems to be much more consistent with focus.  I don't find the files (Mavic 2 Pro) to have the same amount of depth when adjusting in Capture 1 as the X7 but if the shot's not in focus it's worthless.  I plan to visit the only DJI store in the country where I purchased the equipment and get their input, fortunately it's located close to my office.