P4P image not true 20MP

[desert phantom]

Just got my new P4P and have been testing it out.

From what I can gather the camera has a Sony sensor with 20 megapixels.

The image size for the P4P at 3:2 Aspect Ratio: 5472 × 3648 = 20 MP

I extracted the uncorrected image from the RAW/DNG file and it is 5464 x 3620 = 20MP

However, the image has heavy vignetting and significant barrel distortion:

Opening the same DNG file in Adobe Camera RAW, and the image has been corrected. Note that I only opened the file, with the correction seemingly automatically applied (by a profile within the DNG file?)

Note that the image has also been cropped quite abit in the correction process (due to the barrel distortion correction and possibly also for the vignetting). The image size is still 5464 x 3620, so the image has been apparently upsized after the cropping/correction. I.e. it is a 20MP raw image that is cropped down to ~17-18 MP (during the correction process), and then upsized again to 20MP (and hence not true 20MP).

So it seems that the P4P uses a milder version of what GoPro uses to get rectlinear images and video to remove the barrel distortion fish eye look(?).

 

[Jim Roof]

Interesting.

My P4 Pro photos look nothing like yours. I get virtually no barrel distortion and no significant vignetting at all.

It almost looks as if your camera and mine are not the same thing at all.

 

[bill_t]

P4P dng files contain "built-in" corrections that are automatically applied when they are imported into either ACR or Lightroom, and possibly most other raw developer applications They correct for the extreme vignetting on the edges, and the barrel distortion. To be fair, vignetting that extreme would be hard to remove manually. I wish I could find an easy away around those settings and still work with Lightroom, because leaving the barrel distortion uncorrected would make for sharper panoramic stitching. AFAIK there is no way to get around those automatic corrections in either ACR or LR.

desert phantom: how did you extract the uncorrected data file?

Also: does anybody know a raw converter that will import P4P raw/dng files without applying the automatic corrections? Or maybe a way to strip the automatic corrections out of the P4P files?

 

[Jim Roof]

P4P dng files contain "built-in" corrections that are automatically applied when they are imported into either ACR or Lightroom, and possibly most other raw developer applications They correct for the extreme vignetting on the edges, and the barrel distortion. To be fair, vignetting that extreme would be hard to remove manually. I wish I could find an easy away around those settings and still work with Lightroom, because leaving the barrel distortion uncorrected would make for sharper panoramic stitching. AFAIK there is no way to get around those automatic corrections in either ACR or LR.

desert phantom: how did you extract the uncorrected data file?

Also: does anybody know a raw converter that will import P4P raw/dng files without applying the automatic corrections? Or maybe a way to strip the automatic corrections out of the P4P files?

But, my video is also devoid of barrel distortion and vignetting. I have almost zero barrel distortion on my videos and I have made my living as an architectural photographer for the last 35 years and have been dealing with these issues for a long time.

Plus, my live PREVIEW looks just like my photos. No vignetting to speak of and no barrel distortion to speak of.

The OP's images look nothing like what I see in my live view or in my video and/or photos.

 

[Rob8888]

Interesting. I've long suspected this, simply by looking at edge sharpness over the last few months. It's relatively good and a lot better than you'd expect from a camera like this, which made me assume there was considerable cropping in the lens correction profile.

 

[desert phantom]

desert phantom: how did you extract the uncorrected data file?

Just using the windows bare raw converter, which I guess bypasses the dng profile?

But, my video is also devoid of barrel distortion and vignetting. I have almost zero barrel distortion on my videos and I have made my living as an architectural photographer for the last 35 years and have been dealing with these issues for a long time.

Plus, my live PREVIEW looks just like my photos. No vignetting to speak of and no barrel distortion to speak of.

The OP's images look nothing like what I see in my live view or in my video and/or photos.

My live preview, video, JPEG etc. look fine too, i.e. no vignetting or barrel distortion. DJI corrects the output image/video so you don't see any issues. It's only when you bypass this correction process, as above, that you see what the true image, without the corrections applied.

As a photographer myself, I am used to used to the Raw file being very much untouched, so was surprised to see the DNG file with so much correction and cropping etc.

 

[Mindflights]

P4P dng files contain "built-in" corrections that are automatically applied when they are imported into either ACR or Lightroom, and possibly most other raw developer applications They correct for the extreme vignetting on the edges, and the barrel distortion. To be fair, vignetting that extreme would be hard to remove manually. I wish I could find an easy away around those settings and still work with Lightroom, because leaving the barrel distortion uncorrected would make for sharper panoramic stitching. AFAIK there is no way to get around those automatic corrections in either ACR or LR.

desert phantom: how did you extract the uncorrected data file?

Also: does anybody know a raw converter that will import P4P raw/dng files without applying the automatic corrections? Or maybe a way to strip the automatic corrections out of the P4P files?

When I make hdr photos from aeb 3 or 5 dngs, I use Photomatix pro 5 for Mac that don't make that corrections automatically. So you take the 100% of what the sensor saw, as the final result.

After that, I correct it manually in lightroom.

 

[Tomas Wangen]

Interesting find. I replicated the results using the free "UFRaw". Seems the built in profile has quite heavy corrections with some resolution loss; not that surprising given the small optics.

 

[Eddyfisher]

MIB will be coming for those exposing secrets. Best change your name and go into witness protection Phantom.

 

[vidwiz]

I see the vignetting too. The jpg of the same shot is corrected.

 

[Wind_of_Change]

Hey guys, I'm trying to figure this thing out for the P4P, and I'm not sure what's happening. Is the camera saving the source image out to DNG with no corrections? Where are the barrel distortion and vignette corrections taking place? Obviously if RAW and JPG files look identical when I open them in photoshop, then the JPG is being processed on the drone and saved out with corrections. But what about the DNG? Is it also making those changes on the drone when saving the file, or is it saving parameters only that then get read by the RAW importer and applied? If that's the case, then it seems like the raw importer would do a slightly different process than the one performed natively on the drone.

 

[Tomas Wangen]

Hey guys, I'm trying to figure this thing out for the P4P, and I'm not sure what's happening. Is the camera saving the source image out to DNG with no corrections? Where are the barrel distortion and vignette corrections taking place? Obviously if RAW and JPG files look identical when I open them in photoshop, then the JPG is being processed on the drone and saved out with corrections. But what about the DNG? Is it also making those changes on the drone when saving the file, or is it saving parameters only that then get read by the RAW importer and applied? If that's the case, then it seems like the raw importer would do a slightly different process than the one performed natively on the drone.

Opening in e.g. UFRaw will show you the base image without correction. The lenseprofile must be embedded in the dng somehow and is automatically applied by compatible converters.

 

[Wind_of_Change]

I've never used that package, but my morbid curiosity did me in. I downloaded UFRaw and opened some images I'm working on now.

Holy CRAP! My impression of the P4Pro camera has been permanently ruined! It's like waking up with a hangover realizing the woman in the bed next to you looks NOTHING like the hot babe you drunk-flirted with at the pub. (Not that I get drunk and take girls home, I'm happily married, but you get the idea how I feel about it.)

The barrel distortion is huge, the color noise is HUGE, and the vignette is absolutely HUGE HUGE.

Ya gotta hand it to DJI for choosing to automatically perform the processing natively in the drone so that users don't see how bad the source image actually is. I guarantee that the guys that did all the YouTube and web reviews haven't seen these source images or the reviews would have been much different.

After examining some landscape images, I now see where some of the image quality hits are occurring, especially the chromatic aberrations, which appear to be introduced in the automatic post-processing. While they were at it, they should have run some chromatic repairs in their automatic raw process because that's the one thing I always feel a need to fix.

Well honeymoon is over... now I'm way less optimistic about the kind of results I'm going to get using the drone for certain high tech applications. Too bad they can't do a version with a high quality rectilinear lens... oh wait, that would be the inspire 2 with XR5 that costs 4x as much.

 

[AyeYo]

Well honeymoon is over... now I'm way less optimistic about the kind of results I'm going to get using the drone for certain high tech applications. Too bad they can't do a version with a high quality rectilinear lens... oh wait, that would be the inspire 2 with XR5 that costs 4x as much.

One of the things I quickly realized and it makes total sense after seeing this thread, is that stuff like 3D Panos needs to be shot with massive overlap because the distortion is so bad they never stitch well.

 

[Wind_of_Change]

Yeah, photogrammetry work is going to be harder with this camera than I thought when I bought it. My initial tests are giving me a lot higher error rate than I expected, and now I know why. Grr.

 

[AyeYo]

Yeah, photogrammetry work is going to be harder with this camera than I thought when I bought it. My initial tests are giving me a lot higher error rate than I expected, and now I know why. Grr.

Yup. Welcome to my life. "Pro" lulz

 

[Wind_of_Change]

"Pro" lulz indeed. I'm gonna try hard not to complain too much, because I'm still getting a 20mp aerial image, which I couldn't do before in this small of a form factor. It's a HUGE leap ahead from the 12mp gopro stills I've been using. The results aren't stellar, but will still be useful for some of my more creative projects. Here's a little render I did last night with three elements I processed from P4P photos.

 

[Wind_of_Change]

Booo Yaaaaa!

DJI GOGGLES RELEASED!

 

[Meta4]

One of the things I quickly realized and it makes total sense after seeing this thread, is that stuff like 3D Panos needs to be shot with massive overlap because the distortion is so bad they never stitch well.

Yeah, photogrammetry work is going to be harder with this camera than I thought when I bought it. My initial tests are giving me a lot higher error rate than I expected, and now I know why. Grr.

If you are shooting raw for panoramas and photogrammetry, you are fooling yourself.
You are only going to put your images in a blender (stitching program) and no-one is going to see the difference.
Jpgs work just fine for panoramas and photogrammetry.

 

[Wind_of_Change]

Meta4,

I have to respectfully disagree. The quality increase from JPG to RAW is monumental in every way measurable. It's true that JPG images CAN be used, but the question is whether or not they SHOULD be used over the higher quality DNG. If the difference in processing time between them is negligible, why not use higher quality images for a higher quality result? For some applications it's irrelevant, especially if you're going to down-sample your geometry & textures to a game engine or VR use, but for high-end film or even TV FX work, it's definitely noticeable. The JPG noise alone creates a much greater amount of geometry noise when doing photogrammetry, and the same can be said in panoramic shots for subtle gradients like skies.

If your only application in panoramic photography is web delivery, then again, JPG is fine because the format is so small, so you'd be right to use JPG if drive space is an issue. Your very nice gallery is a prime example of this. Your panoramas are beautiful, but they wouldn't hold up as nicely in a large print if they are processed from JPGs, especially those big blue skies. I find that often I'll do a shoot and find other higher resolution applications for the photographs, so it's good to have the DNG on-hand anyway... again, if storage space isn't an issue for you.

I guess it's a question of whether you want to have a professional mentality when you work. I know some amateurs who behave like professionals in their work-flow, and some professionals who behave and think like amateurs in their work-flow.

You could argue that anybody using a P4 Pro, despite it's Pro designation, is an amateur when there are tools like the Inspire 2 and it's higher quality camera to be had. But I can't hike 10 miles through rugged terrain with an inspire and associated gear on my back... heck, I can barely do it with a P4 Pro. So it's a moot argument. You do the best with what tools you have at your disposal and that's all you can do. Which makes my point. Using JPG in no way constitutes "doing my best", especially when the better solution is readily available and takes no more time or effort to use.