The P4P Camera – An attempt to establish The Facts

Joined
Aug 17, 2016
Messages
118
Reaction score
27
A bold title, in order to differentiate this thread from an earlier one, recently closed by the Administrator (quite rightly so). That thread set out to try to discuss whether the advertised spec of the P4P camera was misleading, and unfortunately became rather bogged down by acrimony (though not by the originator)!

This thread is an attempt to discuss the facts about the spec. and performance of the camera, and to mention the possible implications for various users of the P4P, and nothing else! I do not own a P4P (just a lowly P4), but probably will do soon. I am simply interested

So, here are a few things to look over and maybe add some more. Please pick as many holes as you like.

1. The camera incorporates a 1” sensor, make unknown. It’s a legacy description going back to the use of vacuum tube pick-up devices, and they were (perhaps still are?) circular. The actual sensor size is 13.2mm x 8.8mm. It contains 20Mega Pixels (MP), arranged in a 3:2 format.

2. The lens is a sophisticated design, incorporating aspherical components. Is it custom designed or a modified standard design for that sensor size? Perhaps a specialised design has been used, because of the specific needs with the Phantom for a lightweight, high performance lens with minimum field curvature, high resolution and good flare performance .

3. It seems that the lens vignettes at the extreme corners. This means that, in this case, the corners of the image appear black, or darkened. I suppose it could have been heart shaped!

Vignetting can be caused by various things –

a. mechanical, in which a physical object like a lens shade obstructs, this effect softens as the working aperture is widened.

b. optical, when the optics cause an obstruction in the light path. This gets less as you stop the lens down, and

c. sensor, when the pixels at the edge of the sensor receive less light because of the increased incidence angle of the rays.

From the previous thread it maybe that the P4P may exhibits mechanical vignetting. Or maybe a mixture of all three?

4. The vignetting is NOT noticeable in the P4P RAW images because DJI use a preset which crops the 20MP image to around 15MP, which removes the darkening (and the edges of the frame). It is then resized (by the preset) to 20MP. This is a bone of contention, because you then have an upscaled and cropped image. It is possible to view and work on the actual vignetted image using specialist software.

5. The effect described is thought by some to be irrelevant, and by others to be important. Critical stills photographers will expect every portion of their image to be perfect, whereas others may not care about the extreme corners. For those using aspect ratios wider than the native 3:2 it will be unimportant. Video users, for example, will normally work with 16:9 aspect and see crisp, clean images, I expect.

oooooooOOOOOOOOOOOoooooooo

It would be good, if you do want to add to this information, to quote the numbers above, or add your own for points I’ve missed.

In any event, I hope this gives a little light reading and maybe helps to clarify things slightly.
 
Ok, so they are cropping down and re-zing up......get it. My P3P wide video is soft around the edges so I guess the crop to 35mm format......the stills are extremely impressive. If the P4 is as good, i wouldn't complain. But I do see your point, vignette is much easier to correct with Lightroom than soft corners...!
 
Last edited:
  • Like
Reactions: Geoff G
Yes, the sensor on P3P looks like the same spec. as the P4 standard and I'm very pleased with mine.

Must be useful to have adjustable aperture and mechanical shutter on the P4P though! In the UK the ND filters are on and off like .........well, quite often!
 
1. I don't think there is any question it is a 1" sensor with 3:2 aspect ratio (nice bit of trivia some might not have known re the relationship of sensor size to a 1" thermionic imaging tube).

2. I suspect all that might be said about the lens is that it is small and light and likely very cheap to manufacture. Resolution appears good at the centre, curvature correction seems poor with average flare performance.

3.
a. No- unlikely

b. Almost certainly- all optical lenses produce less brightness at the edges of the circle of illumination.

c. Probably- will be of less significance than lens contribution. .

4. Seems to be the case and universally agreed.

5. The subject issues must be less significant at 16:9 vs 3:2 as less if the sensor corners are used to render the image readout.
 
  • Like
Reactions: AyeYo
1, we were told by the reviewers and DJI themselves in the press releases that it's a Sony sensor
2, It's an 8mm lens - of course there is barrel distortion (curvature)
3, yes, it does suffer from vignetting as most 'fisheye' type lenses do
4, wrong - the vignetting is visible in the raw images (no need for caps as it's not an acronym) - it's just that DJI have chosen to hide this with an embedded profile that 'fools' the most commonly used raw editor (Adobe Camera Raw). The argument is that the idea of a 'one size fits all' profile is no better than just restricting the P4P to jpeg output. A raw file should be just that, a raw file, exactly what the camera 'saw'. The embedded profile performs a series of 'destructive' edits including cropping the image down to approx 15mp before re-scaling to the claimed 20mp output.
5, It's important to some people because they submit images to print houses who will very often reject interpolated/enlarged images - if all you do is look at stuff on your screen or upload images to facebook it really doesn't make any difference.

All of this was discussed in the previous and other threads - so why start over again?
 
  • Like
Reactions: AyeYo
I'm curious, why do you consider it to be a fisheye type lens?
It's not a fisheye lens but he's used the term loosely when what he means is that it's a wideangle lens with some barrel distortion (not an uncommon thing in wideangle lens designs).
That distortion is corrected out in the lens profile that's been mentioned but is visible if you use software to see the pre-corrected image.
 
So...how do I go about getting the 'full' raw file before the degenerative profile is applied and then is it possible to work with it in Lightroom?
 
So...how do I go about getting the 'full' raw file before the degenerative profile is applied and then is it possible to work with it in Lightroom?

The easiest way is is to use a different raw editor - I prefer Rawtherapee - it's free and integrates with Lightroom/Photoshop/Adobe Camera Raw. Others will suggest other options, it's down to personal opinions
 
  • Like
Reactions: whitelineNI
I'm curious, why do you consider it to be a fisheye type lens?

I put 'fisheye' in inverted commas deliberately.

It's an 8mm lens (anything below 10mm would normally be classed as a ultra wide angle or fisheye) so call it what you will, it is prone to quite severe barrel distortion (just like you would expect from a fisheye)

We've done all this already on several threads and it never really goes anywhere
 
  • Like
Reactions: AyeYo
I put 'fisheye' in inverted commas deliberately.

It's an 8mm lens (anything below 10mm would normally be classed as a ultra wide angle or fisheye) so call it what you will, it is prone to quite severe barrel distortion (just like you would expect from a fisheye)

We've done all this already on several threads and it never really goes anywhere
Your 8-10mm fish eye/extreme wide angle classification has widely accepted application to full frame 35mm imaging systems. It can't be universally applied to other formats. In evaluating the distortion of the lens to the extent we might choose to compare it to full frame DSLR we would say it's a wide angle.

As you have said there is nothing that might be gained from this thread that hasn't been covered. The subject observations are well established. DJI has implemented on board processing to attain the best overall performance that might be expected from the employed optics where it would seem weight and very likely production costs are the primary design criteria.
 
  • Like
Reactions: DaRana and Meta4
Ian, how certain are you this is the sensor model?

If I choose to shoot 16:9, will it still apply crop&upscale?

I always crop my shots to 16:9 anyways and then scale them down to 4K resolution. This way I could have the same image that wasn't upscaled to begin with.
 
This same CMOS chip is used in Astrophotography, nice sensitive chip.
 
I don't believe the corner darkness to be vignetting in the traditional sense. The darkening is sudden and abrupt and is accompanied be defined band of image degradation in my case. This is evidence to me of "image circle fall off" (ie the lens does not produce an image circle large enough to cover the entire sensor). This is most noticeable on mine with it's slightly "off centre" lens, as pulling 5 stops will still not expose any captured image in the extreme corner.

With that said, I think it best that everybody simply knows that DJI's 20mp image is in fact an interpolated 20mp image and not a "sensor size" 20mp. From there they can make their own informed decisions as to whether the camera is suitable for their needs.
 
I don't believe the corner darkness to be vignetting in the traditional sense. The darkening is sudden and abrupt and is accompanied be defined band of image degradation in my case. This is evidence to me of "image circle fall off" (ie the lens does not produce an image circle large enough to cover the entire sensor). This is most noticeable on mine with it's slightly "off centre" lens, as pulling 5 stops will still not expose any captured image in the extreme corner.

With that said, I think it best that everybody simply knows that DJI's 20mp image is in fact an interpolated 20mp image and not a "sensor size" 20mp. From there they can make their own informed decisions as to whether the camera is suitable for their needs.

Mine is very different to that and with a little work the corner detail is possible to recover so it's certainly being recorded - I've used this image before but it gives a decent illustration
RT.jpg
 
Mine is very different to that and with a little work the corner detail is possible to recover so it's certainly being recorded - I've used this image before but it gives a decent illustrationView attachment 86410
It would be interesting to learn how many MP the resultant edit would be with the dark corners removed (cropped out). It is very evident why DJI elected to apply in camera pre-processing to the RAW. The lens is unable to acceptably illuminate the 1" sensor. The end result is still better than you might get from any other RTf package anywhere near the price point.
 
It would be interesting to learn how many MP the resultant edit would be with the dark corners removed (cropped out). It is very evident why DJI elected to apply in camera pre-processing to the RAW. The lens is unable to acceptably illuminate the 1" sensor. The end result is still better than you might get from any other RTf package anywhere near the price point.

the red line roughly shows what the crop is (about 14.5mp) but (if you'd seen the other threads) that is only a small part of the story

rt2.jpg


The embedded profile also adjusts the white balance, boosts highlights, corrects the barrel distortion, supposedly corrects any chromatic aberration then re-scales the image back up to 20mp. These are 'destructive' edits if you use Adobe Camera Raw as there is no way you can reverse the process.

Here's what Adobe Camera Raw sees (all sliders set to zero). As you can see the shadows are way too black, whites and highlights are blown, reds way over saturated, contrast has been boosted and edge detail is much softer than in the raw file viewed in Rawtherapee. This is not what you expect of a raw file.

A quick look at the Adobe raw should tell you that there is a lot wrong - The length of the shadows suggest it was taken early morning (or late evening) so the light should be 'soft' and have a golden hue to it - exactly how it looks in the original unaltered raw. In this version it looks like midday and that's hopeless.

ACR.jpg


It doesn't matter what it compares to and how much of a bargain it is (strawman/fanboy argument) what matters is that it's not outputting what the specifications say and to some of us that matters. I can't say for sure but it's possible that if I had been aware of this issue I would have probably looked at a more expensive alternative platform - every single image being a compromise is not what I call a 'professional' piece of kit.

I said in one of the other threads that publishers and print houses don't want/refuse interpolated images so without a lot more work than should be required the P4Pro struggles to produce the kind of images that it should be capable of and that we were led to believe when we bought the product before this 'cheat' was known
 

Members online

Forum statistics

Threads
143,066
Messages
1,467,356
Members
104,934
Latest member
jody.paugh@fullerandsons.