DNG RAW Make A Difference on Maps & 3D?

Consultant · Oct 10, 2017

Has anyone compared orthomosaic and 3D model results (from the likes of DroneDeploy, Pix4D, etc etc) between using .dng RAW files and .jpg JPEGs to see if there's any noticeable difference in the output? There a lot larger storage requirement and upload/download time for the dng compared to jpg so if there's not a significant difference in the finished product then it would seem dng is a waste of space/bandwidth. I'd be most interested to hear experience from people with the larger sensor camera on the Phantom 4 Advanced or Phantom 4 Pro but the P4A or P3Pro would be interesting too.

It's one thing to theoretically claim it makes a difference but who's actually compared?

Meta4 · Oct 10, 2017

Consultant said:
Has anyone compared orthomosaic and 3D model results (from the likes of DroneDeploy, Pix4D, etc etc) between using .dng RAW files and .jpg JPEGs to see if there's any noticeable difference in the output? There a lot larger storage requirement and upload/download time for the dng compared to jpg so if there's not a significant difference in the finished product then it would seem dng is a waste of space/bandwidth. I'd be most interested to hear experience from people with the larger sensor camera on the Phantom 4 Advanced or Phantom 4 Pro but the P4A or P3Pro would be interesting too.

It's one thing to theoretically claim it makes a difference but who's actually compared?

Your guess is right.
If you are taking a large number of images with big overlaps and then "putting them in a blender" to create a huge orthophoto composite image, you are probably fooling yourself.
I haven't tried to compare because DroneDeploy only shoots jpg and I suspect the camera would have problems writing so many big files to the sd card during the mission.
In general the idea that raw is the only image format anyone should consider is exaggerated and for most people, most of the time shooting jpg is all they will need.
maybe it was a valid argument in the days of earlier digital cameras but these days it's not so important.
The P4 pro produces excellent quality jpg files.

JGutz138 · Dec 2, 2017

I know this is an older question but thought I'd share my experience since it seems I'm the only person that's taken the time to do a comparison. I've ran this test with my P3P and then again with my Inspire 1 pro with the x5 camera and 12mm oly, both with the same results... I found minimal differences between the final 3d models and orthomosaics between jpg and raw data sets. Processing time and error residuals were almost identical. A few things to note though is that there isn't a single photogrammetry software that I found that can process raw dng files, so a conversion to another format is needed right off the bat(I used agisoft photoscan for the test and they say that it prefers uncompressed tiff files so that's what I used). Also dng images look flat and unsaturated when compared to the jpg and the real world so they also require some post work before being used for orthomosaic generation to get a good looking ortho. The only advantage I really found is that the added dynamic range of the dng images leaves a lot of room for any post adjustments that may need to be made...for example during one of the test flights there was intermittent cloud cover so some images in the data set were underexposed when cloud coverage happened, I was easily able to adjust the underexposed dng images without introducing alot of artifacts in post and that was not the case with the jpgs...the jpgs required a second flight to have a usable dataset. Also I did notice an improvement in areas with limited color variation like black pavement (photogrammetry software has a hard time processing accurate 3d data in areas like this) as there was a noticeable amount of more variation in the dng files. The worst part about trying to shoot in dng though is that almost no data acquisition apps(pix4d mapper, drone deploy, altizure) support it and even when you find something that does(UgCS), dji has implemented a required 10 second delay between raw dng captures even though it only takes less than 4 seconds to write the file to the sd card so flight times at least triple. So all in all shooting in dng could potentially save some headaches with the added post processing ability but at the same time it introduces many more steps and headaches into the process. That being said, I still shoot some flights in dng, mostly when I have to be up above 250' AGL and if I could find a way around the 10 second delay I would shoot every flight in dng and then export them to larger jpgs after post and just delete the dng files. I found overlap settings, GCP placement and following terrain elevation to be much more critical when 3d accuracy is important.