How to take HDR 360 with AEB automatically

[Andrex]

How long does this entire process take to get one usable finished pano? 5 hours? Is it really even noticeably better than the 60 seconds from start to finish that the M2P takes, including cloning in a ceiling/sky top? I have shot as many as 12 different panos over a 3 mile area in a single M2P flight. I do save the original 26 jpg images on each for post processing into a 75MP stitch, only on the best ones. However, when 95% of your viewers are viewing the pano on a smartphone, even it is overkill, and delays loading the extra detail, except for my own viewing pleasure on a desktop. Perfection is overrated. Sometimes good enough is good enough!

Give it a rest. It won't take 5 hours! My process is just over an hour and I'm sure I can build efficiencies into that. If your computer is relatively decent, processing 170 photos shouldn't be too onerous.

We're interested in producing even better high-quality 360s, not just settling for OK ones. Take this as an example (Google Maps). It's good, it's been post-processed and stitched on a computer, but it isn't HDR and could have been much better. I'm willing to put a bit more time into that, even if not everyone is.

Are you interested in sharing some of your 360 panos, out of interest?

 

[kawartha]

Thanks for that @kawartha, I have made a note of it and will give it a try. I was also wondering if you have done any 180 panos and if so what settings you would recommend. I have tried 3 rows by 6 columns a few times and that works OK but would love to hear how others go about it.

Yes, I've done 180 as well, also 150 and 120. Usually 6 columns and 2 rows at 45-60 degrees gets all you need for a 180. Lately though I just shoot a 360 and if I want only 180 or less then I just pick the ones I want from the processed HDR. However I am concerned about disk space (this is why I do the panos in JPEG), so I'll probably be more careful in the future.
It does not take 5 hours to get one processed pano. One M2P battery will get you at least 3 panos, more if you are doing 180 or less. I'm using the standalone Photomatix Pro batch mode, don't use Lightroom. Sometimes I process one representative image and tweak the settings and then save them as a preset. Processing 170 images for 5-bracket HDR takes less then 10 minutes. PTGui is a couple of minutes, unless there are alignment problems, then it can be a while. I am not too good at fixing alignment in PTGui right now. I don't have any online 360 examples to show you but here are a few 150-180 degree examples (I've downsized them all to take up less space here):

Width: 150 degrees, Columns: 5, Rows: 2, Height: 45 degrees


Width: 180 degrees, Columns: 7, Rows: 2, Height: 40 degrees



Width: 180 degrees, Columns: 6, Rows: 3, Height: 60 degrees. This one has 2 alignment errors on the horizon I could not fix.

 

[Deleted member 94047]

Width: 150 degrees, Columns: 5, Rows: 2, Height: 45 degrees

I really like how this came out. Will give it a try. Many thanks!

 

[GadgetGuy]

Give it a rest. It won't take 5 hours! My process is just over an hour and I'm sure I can build efficiencies into that. If your computer is relatively decent, processing 170 photos shouldn't be too onerous.

We're interested in producing even better high-quality 360s, not just settling for OK ones. Take this as an example (Google Maps). It's good, it's been post-processed and stitched on a computer, but it isn't HDR and could have been much better. I'm willing to put a bit more time into that, even if not everyone is.

The primary reason for using HDR is to bring back detail in otherwise overexposed skies. Properly exposed skies can easily be cloned in in post, and you have to anyway, because no DJI drone gimbal can elevate above 30° above horizontal. I would recommend using your own library of properly exposed sky images, and adding them to the automated M2P Spherical Panos of your choice. Comparing 60 seconds in flight with an already cloned in ceiling vs. a full hour of post processing plus shooting time, is at least is 60x longer, for what is likely to only be perceived as a marginally better result.

Perhaps I should be asking exactly what you intend to do with the end result of all this extra work, and on what type of device your viewers will likely consume the result. If is only a Smartphone, I doubt the difference will be noticeable.

 

[Deleted member 94047]

Seriously @GadgetGuy, give it a rest. It is not like he (or anyone else who chooses to do this including me) is taking those precious hours from you. Yes, it takes a whole lot of time compared to the available auto Pano modes. Is the end result worth the extra time? I believe that question is best left for the person who chooses to do it. For me, yes it is. Also, what are you on about HDR being primarily useful for "bring[ing] back detail in otherwise overexposed skies". That is certainly one of its uses but it far from the only one. HDR is extremely useful in any environment where you have a mixture of shadows and bright light - even when you are shooting 90 degrees down and there is no sky in your shot. In fact, I shoot almost everything in AEB these days. If I think the extra step of merging the set into an HDR image is not warranted or worth the time, I just take the middle exposure and develop that.

 

[GadgetGuy]

Seriously @GadgetGuy, give it a rest. It is not like he (or anyone else who chooses to do this including me) is taking those precious hours from you. Yes, it takes a whole lot of time compared to the available auto Pano modes. Is the end result worth the extra time? I believe that question is best left for the person who chooses to do it. For me, yes it is. Also, what are you on about HDR being primarily useful for "bring[ing] back detail in otherwise overexposed skies". That is certainly one of its uses but it far from the only one. HDR is extremely useful in any environment where you have a mixture of shadows and bright light - even when you are shooting 90 degrees down and there is no sky in your shot. In fact, I shoot almost everything in AEB these days. If I think the extra step of merging the set into an HDR image is not warranted or worth the time, I just take the middle exposure and develop that.

Personally, I find HDR images unrealistic and lacking in contrast. The sky reference was taken directly from one of the posts above. Until the final purpose of the output is known, and what devices(s) it will be viewed on, which I will allow the OP to answer himself, it remains to be seen whether it is, in fact, worth it, other than just to the pilot. The OP keeps referring to "we" so the questions are relevant. Shooting the pano in even lighting and getting it on the take, has always made more sense to me than trying to "fix it in post." YMMV. The exact same pano shot from the same location will look very different on different days and at different times of the day and night. Rather than trying to get it all in one attempt, I would rather shoot it multiple times on different days in different lighting, and pick or tweak the best ones. Each to their own.

 

[Andrex]

Two attempts from today with the Mavic 2 Pro. Note the changing conditions (hence the nice rainbow). First one was taken using DJI Go app with a standard spherical pano of 26 DNG images. Processed these in Adobe Camera Raw, then applied lens correction in Photoshop, stitched in PTGui and used Photoshop's content-aware fill to finish the sky.

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

goo.gl

The second was taken using Litchi. Five-bracket AEB, pano on settings to take around 25 photos. I then loaded the DNG files straight into PTGui and it detected the AEB images, merged them, before I applied some tone mapping and stitched. Thereafter the process was the same as #1, with Photoshop content-aware fill completing the sky. I don't know why, but although the rainbow is there, it seems to have had the colour sapped from it, whilst I have actually boosted the colour/contrast in the rest of the photo...

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

goo.gl

For now the first workflow seems to produce better results. I may consider using Photomatix if there is a trial period. I may need to take more time and care on the second workflow to improve the result, and it seems more de-ghosting is required...?

 

[GadgetGuy]

We're interested in producing even better high-quality 360s, not just settling for OK ones.

May I ask who the "we" is that is tasked with the job of producing the 360s, and for whom the 360s are being created, for what purpose, and the types of devices upon which the 360s will be viewed by the end user?

I'll be happy to share some of my 360's with you, once I know more about your commercial needs for the output, assuming I haven't misinterpreted the "we" reference.

 

[Andrex]

May I ask who the "we" is that is tasked with the job of producing the 360s, and for whom the 360s are being created, for what purpose, and the types of devices upon which the 360s will be viewed by the end user?

I'll be happy to share some of my 360's with you, once I know more about your commercial needs for the output, assuming I haven't misinterpreted the "we" reference.

The 'we' refers to myself and the other posters in this thread who have also either expressed an interest or indeed previously carried out HDR 360s.

I don't know how the 360s will be consumed as they are often in the public domain. I prefer to view on my laptop for a fuller experience. What I can say with confidence is that they won't be viewed on a piece of photo paper...

Can we please focus on technique here? That's the point of this sub forum.

 

[GadgetGuy]

The 'we' refers to myself and the other posters in this thread who have also either expressed an interest or indeed previously carried out HDR 360s.

I don't know how the 360s will be consumed as they are often in the public domain. I prefer to view on my laptop for a fuller experience. What I can say with confidence is that they won't be viewed on a piece of photo paler...

Can we please focus on technique here? That's the point of this sub forum.

Thanks for clarifying. I misunderstood the "we".
Lots of commercial applications for 360s where efficiencies matter, as well as the primary device used for consumption, which today is, unfortunately, usually a smartphone on a cellular network, where the file size of the image matters, as higher detail slows image loading, and consumes more cellular data, and the benefits of HDR may be completely lost. It's no different than creating images for web use. Smaller file sizes are favored, which is why DJI settled on a 12MB output file, even though a 75MB version can be created from the original 26 images. As you have experienced, the longer the shooting process from AEB, the more changes happen within the scene being photographed. Like most things, it is a series of compromises at every step.

In terms of technique, I have found that varying the starting direction has the greatest impact on the end result. Pick a direction to start from that has a neutral exposure relative to the rest of the scene. Vary the starting direction to see the difference. Lighting is also critical. Flat lighting on cloudy days produces the best panos. If the sun is directly above, it won't blow out one side, but the harsh lighting isn't as appealing as a giant softbox from overcast high clouds, with clear air for distance visibility. Getting it on the take can avoid lots of fixing in post.

I commend you for your efforts. When you are only shooting for yourself, you have the luxury of trying for perfection. Keep at it!

 

[Andrex]

Haven't had much of a chance to go out and try more techniques yet, so just posting this in the meantime to demonstrate that HDR isn't necessarily required every time.

This pano was shot using DJI Go and processed only from the raw DNG files using Photoshop and PTGui. The lighting and cloudy skies mean that not a lot of detail was lost in the sky. To the east you can see the stormy sky which had passed through only 5-10 minutes previously.

https://goo.gl/maps/AaEqmdV2BVBnEyjE8

That said, there is still approx. 10% of the sky which is blown out and irrecoverable even from the RAW files which may have benefitted from HDR. In a scene like this, 'faking' the sky wouldn't work as well, as capturing the genuine weather was a part of the shot.

For info, this is Ironbridge in Shropshire, where the flood defences were up on the River Severn which peaked at 6.2 m at the exact time of this photograph.

Thanks for clarifying. I misunderstood the "we".

Yes you did, but I appreciate the post this time.

I did the above photo only for myself and the users of Google Maps, but I'm really pleased with it, because it shows not only a landscape but an 'event' (the river peaking and flood barriers in action).

 

[vili]

Just set it so there’s a little bit of time between captures so the SD card can digest the three exposures before going to the next one.

Man, this is the MOST IMPORTANT THING regarding AEB Panos.
If i don't add extra time before/after each aeb sequence, shutter speed info will be often wrong in the exif data on my mavic air.
Yesterday i tried with 2-2 sec, and all photo's exif is perfect!!

This is a secret, dont tell to anybody else

 

[ManuFly]

Haven't had much of a chance to go out and try more techniques yet, so just posting this in the meantime to demonstrate that HDR isn't necessarily required every time.

This pano was shot using DJI Go and processed only from the raw DNG files using Photoshop and PTGui. The lighting and cloudy skies mean that not a lot of detail was lost in the sky. To the east you can see the stormy sky which had passed through only 5-10 minutes previously.

https://goo.gl/maps/AaEqmdV2BVBnEyjE8

That said, there is still approx. 10% of the sky which is blown out and irrecoverable even from the RAW files which may have benefitted from HDR. In a scene like this, 'faking' the sky wouldn't work as well, as capturing the genuine weather was a part of the shot.

For info, this is Ironbridge in Shropshire, where the flood defences were up on the River Severn which peaked at 6.2 m at the exact time of this photograph.



Yes you did, but I appreciate the post this time.

I did the above photo only for myself and the users of Google Maps, but I'm really pleased with it, because it shows not only a landscape but an 'event' (the river peaking and flood barriers in action).

Hi, I congratulate you for your results.
Regarding the 360 hdr panoramas I wanted to report this App.
hdrpano iOS APP
Never used?

 

[vili]

Only iOS
Is it better than Litchi for aeb panos?

 

[ManuFly]

Only iOS
Is it better than Litchi for aeb panos?

Yes, it is only for iOS.
I haven't had the pleasure of trying it because of coronavirus yet.
I have seen the youtube channel and it seems to me very well.
When you can fly I would like to shoot both pano hdr 360 and pano 180 hdr.

 

[Andrex]

Hi, I congratulate you for your results.
Regarding the 360 hdr panoramas I wanted to report this App.
hdrpano iOS APP
Never used?

Not possible to use this, since it is not on Android.

 

[dhpdigital]

Hi. Thanks to everybody who has contributed to this thread. I have a M2P and found this forum whilst searching to see if this was possible. I capture 360 32bit hdri’s on the ground but wanted to know if I could do the same in the air.
My reason to do this is to create backplates and environment maps used in CGI renders. So the only way to do this is to capture RAW, stitch them, and tone map them. To get the best most professional image the hdri is the only way to go IMO. I get that the look can be a bit flat or lacking contrast, but if you know what you’re doing it’s just a case of applying the right technique for the tone map.
If you don’t know about it and want to, this is an amazing resource Overview

Kind regards, D

 

[Andrex]

HDR 360 panorama example: Google Maps

1. Use Litchi, shoot in manual mode, capture 5 AEB photos and switch to Panorama mode.
2. Take 200 RAW photos and carry out lens correction in Photoshop (generates 20 c. 100 mb TIFF files)
3. Add TIFF files to PTGui Pro and enable HDR mode. Tone map, and stitch.
4. Open stitched file in Photoshop and fill in the sky using content aware fill, making some other subtle corrections as required.

 

[Andrex]

Another example from The Wrekin in Shropshire - I feel like the clouds and sunshine / shadows really enhance the dynamic range in this 360. I'm starting to get much more comfortable with generating 360 HDR panoramas now. I should really have done this shot about two weeks' ago to get the oil seed rape fields at their maximum bloom...

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

goo.gl

 

[Andrex]

Revisiting this almost a year on (because winter is when you tend to need to use HDR more to beat the low sun) this is today's shot from Shrewsbury in flood.

Google Maps

Find local businesses, view maps and get driving directions in Google Maps.

goo.gl

Unfortunately using a 5-bracket AEB in Litchi means the 34-shot panorama took some time. What seems to have happened in the 5-10 minutes it took to complete a circle is the drone drifted in the wind (slightly) and also appears to have lost 10 m in altitude, which I didn't notice during shooting. The panorama started at c. 75 m above ground.

The result is unfortunately an issue in stitching the start and end seam of the panorama (parallax, and also a lack of sufficient overlap, but I think mainly paralax). Disappointingly, this is most pronounced for two of the town's spires - in the opening shot, the spires were below the horizon, but come the final shot the spire tip was above the horizon.

I attempted to manually fix in photoshop with clone tool, dodge, burn and the original frames, but I have struggled.

Apart from one slither of the photo representing probably 5 degrees, I'm happy with the rest. But a five degree aberration is pretty significant in a photo like this.

Lessons? Not sure. The naked eye probably wouldn't pick up on the potential parallax over 10 minutes of auto-pano shooting in particular between the start and end frames. A drone can drift a lot in 10 minutes. Better keep an eye on the altitude reading next time I guess. I don't think anything more can be done in post-processing to save it.