Poor quality raw 48MP Mavic Air 2

[brett8883]

@brett8883 thank you for your question, but not, my friend shoots in raw, than he edits in PS/LR, as me. I'm happy to hear that my 48MP are perfect for you, I'll post here some details in order to make you see the disturb I noted.

View attachment 131463

View attachment 131464

View attachment 131465

@sar104 thank you for your opinion, I answered my friend to send me a pair of DNG by his Mini 2, in order to make the comparison, I'm waiting for him; once I received em, I'll post here!

So let me ask you, what were you expecting to see when you zoomed in 10x like that? What you should expect to see is a pixilated low resolution image which is what you got. You are zoomed in to 0.4MP see below.
This is how far you can zoom in and still have as much resolution as the Mini 2. You were zoomed in many times more than that.
Any photo you zoom that far in will look screwed up.

Here's a photo I took will my 61MP Sony A7RIV with the 135 f/1.8 G Master which is a much larger sensor with far better opticsView attachment A7R05414.jpg)



If I zoom into 0.4 MP it looks terrible too. Thats just how it works. You can't zoom in infinitely with any camera

 

[ilSimo95]

Well, I think you don't understand the problem. I'm not talking about quality loss, which is simply natural cropping an image, but the point is: I'm cropping to show you the noise, which all my friends (drone operators) saw, I don't know why in these forums everything has to be seen every time as if everything is fine.

The image you posted here is great, ok, and cutting it loses the quality as natural, but there is no disturbance!

I will post the photo my friend sand me there, and it has already been cropped vertically. If you don't see the differences here, I'll give up.

 

[brett8883]

Well, I think you don't understand the problem. I'm not talking about quality loss, which is simply natural cropping an image, but the point is: I'm cropping to show you the noise, which all my friends (drone operators) saw, I don't know why in these forums everything has to be seen every time as if everything is fine.

The image you posted here is great, ok, and cutting it loses the quality as natural, but there is no disturbance!

I will post the photo my friend sand me there, and it has already been cropped vertically. If you don't see the differences here, I'll give up.

Higher resolution sensors have more pixel noise because fewer photons hit each individual pixel for a given exposure duration. However, the overall amount of noise is the same because there are far more pixels and it gets averaged out over more pixels. When you zoom in to pixel level you will see more noise on a photo taken with a high resolution sensor than a lower resolution sensor. This is also why low light cameras will have lower resolution sensors because larger pixels=less noise per pixel. This is always a trade off but with the quad bayer sensor on the Mavic Air 2 you get mostly the best of both worlds because for better noise capability you switch to 12 MP mode which acts like a 12 MP sensor with larger pixels. That’s the whole point of quad bayer.

The “disturbance” you see, I see it too it’s not like I don’t see it, is due to this AND artifacts from the unique debayer algorithm used to debayer the quad bayer sensor.

I see what you are seeing too, however, what I am trying to tell you is look how far you have to zoom in to see it. Ask yourself does it really matter if you’d never crop that far in on any camera?

Your images images look great and I don’t really need to see anybody else’s photos to establish that.

 

[sar104]

I have the air 2 with a few hundred flights so far.the 48mp setting is a gimic and the camera on it is a true 12mp camera.I know it says 48mp on the camera face which should really never have been put on there.
I never use the 48mp trickery setting as the photos do not look good.
Just my 2 cents worth.

Your two cents are incorrect. While you may not subjectively like the results, that is a 48 MP sensor.

 

[Wreckhunter]

Your two cents are incorrect. While you may not subjectively like the results, that is a 48 MP sensor.

Your two cents are incorrect. While you may not subjectively like the results, that is a 48 MP sensor.

Did you look in the user manual you will see 12/48 enough said.

 

[sar104]

Did you look in the user manual you will see 12/48 enough said.

Yes, but you don't even need the manual - you can easily look up the specifications and design of the sensor, which is a Sony IMX586. And the 12/48 refers to the available images dimensions, not the sensor array size.

 

[sar104]

Yes I am aware of the sony IMX586.
The 12 megapixel sensor,I repeat 12 megapixel sensor utilizes Quad Bayer technology to split the readout of the image,effectively quadrupling its image.
A question, do you own the mavic air 2,just wondering.
Go around and check the articles and you will find it is in fact a true 12 meg.sensor period.
If it was a true 48 megapixel sensor in the mavic air 2 it would sabotage mavic 2 pro sales,and I know
DJI is smarter than that.
Besides getting a true 48 megapixel sensor out of only a 1/2 inch sensor would be difficult.
I photograph shipwrecks,kinda know a little about my drone.
Some android smartphones use the same quad Bayer technology to make it look like it has a higher megapixel count.
Not trying to be smart ,just the way it is.
The End

Which illustrates that using a drone to photograph shipwrecks, or any other subject for that matter, doesn't teach you anything about how the drone or its camera works. And your concern over whether I own a Mavic Air 2 is equally misplaced, since it is irrelevant to my understanding of optics and sensors, which comes from being a physicist. No - I don't have a Mavic Air 2, but if it makes a difference to you I do have a Mavic 2 Enterprise Advanced, which has the same 48 MP sensor on the visible imaging side of the camera.

Anyway - repeating your misunderstanding of how a Quad Bayer sensor works doesn't make it correct. The base pixel count of a sensor refers to the number of photosites on the sensor. A 12 MP Bayer sensor has 12 million photosites (3 million red, 3 million blue, 6 million green). A 48 MP Quad Bayer (such as the IMX586) has 48 million photosites (12 million red, 12 million blue, 24 million green) - exactly the same as a 48 MP regular Bayer sensor but with a different arrangement of the color filters.



A Bayer sensor (quad or regular) has luminance data from every sensor, which determines its spatial resolution, but always has fewer photosites for each color than final pixels since each final pixel needs 3-color data. Those color data are interpolated from neighboring photosites via the demosaicing algorithm, which varies by photosite arrangement.

The reason for the quad design is to allow it to run in more than one mode. It can demosaic at every photosite to get a 48 MP image, it can bin each group of 4 similarly filtered photosites to reduce noise giving a 12 MP image, or it can set the gain or integration time differently within a group to get 12 MP HDR image from a single exposure.

So when run in full demosaicing mode the Quad Bayer spatially has a full 48 MP spatial resolution, and that's apparent in the fine detail, but it has to do more interpolation to resolve the colors properly across all the pixels and will generally have poorer color detail resolution than a 48 MP regular Bayer. But it is still a 48 MP sensor.

 

[sar104]

Im sorry but if you feel you have won this and if it make you feel better so be it.
The sensor in the camera is 12 megapixel as I have said before.
And I will no longer answer this thread as i have better things to do.

It would have been better if you had refrained from wasting your time and mine by not posting in the first place.

 

[sar104]

I agree with you 100%,they should have left it at what it really is 12 mp CMOS sensor period.
And left out the Quad Bayer technology or trickery I should say.
A true 48 mp sensor currently ,or should I say the camera size required currently
to house that sensor I think would be to big to fit in a folding drone with current
technology.I may be wrong,but it would make perfect sense.
Have a good day.

Are you incapable of understanding, or just unwilling to look at, the specification sheets for this sensor? No - the physical sensor size depends on the photosite size as well as the number of photosites. The IMX586 uses 0.8 µm photosites to put 48 million of them on a 1/2 in. sensor.

 

[Wreckhunter]

I've done a lot of testing, and 48MP doesn't really offer more effective resolution than the 12MP mode. It's basically interpolated just to get to 48MP (even though it technically isn't).

If you want the resolution, you're better shooting 12MP raws, then using the super resolution feature in Lightroom/Photoshop to upscale the raw file. It's a cleaner result, with fewer artifacts.

I went to the Lake two days ago with my air 2 and shot stills in 12 and the 48 mp setting and found no difference at all.
This was just my latest test with the 48 mp setting.
They should have left the camera at just 12 mp which is what the honest mp count is anyway.
I know some argue that this sensor on the air 2 is an honest 48mp sensor.
which it really is not.
My wife processed the photos for me ,as she has a degree in photo work.
She could not tell the difference between the 12 or 48mp setting.
Quad Bayer technology is crap to make a long story short.
Have a good day.

 

[sar104]

I went to the Lake two days ago with my air 2 and shot stills in 12 and the 48 mp setting and found no difference at all.

So the images had the same number of pixels, independent of whether the camera was set to 12 or 48 MP?

 

[Wreckhunter]

So the images had the same number of pixels, independent of whether the camera was set to 12 or 48 MP?

Right

 

[sar104]

Right

No, not right. Apparently you don't understand the camera settings. A 48 MP image from that camera will be 8000 x 6000 px, and plenty of examples have been posted here and elsewhere. A 12 MP image will be 4000 x 3000 px. It's not very difficult to tell the difference.

 

[Wreckhunter]

No, not right. Apparently you don't understand the camera settings. A 48 MP image from that camera will be 8000 x 6000 px, and plenty of examples have been posted here and elsewhere. A 12 MP image will be 4000 x 3000 px. It's not very difficult to tell the difference.

I understand fully the camera settings.

 

[Meta4]

I understand fully the camera settings.

Apparently not

 

[sar104]

I understand fully the camera settings.

This is a good example of pointless, ignorant posting. This isn't a matter of opinion; it is a clearly demonstrable fact that the camera will produce either 4000 x 3000 (12 MP) or 8000 x 6000 (48 MP) images.

And the number of photosites on the sensor and the number of pixels in the image are unrelated to whether the sensor is a Bayer or Quad Bayer - that simply changes the color demosaicing method used and arguably the color spatial resolution. Quad Bayer and regular Bayer filters have different advantages, and the reasons for using a Quad Bayer rather than regular Bayer filter with this sensor are obvious, as I pointed out in post #27.

You can certainly argue about IQ differences between equivalent-sized regular Bayer and Quad Bayer sensors at full resolution or about the IQ differences between the 12 MP and 48 MP images from this sensor, but disputing the physical specifications of the sensor is just fatuous. And if you can't even figure out either how to set it to take 48 MP images, or if you can't figure out the difference between 12 MP and 48 MP by the pixel dimensions of the resulting images, then you need to start by learning some digital photography basics.

 

[Wreckhunter]

Apparently not

Then what is your solution if I may ask.

 

[Meta4]

Then what is your solution if I may ask.

I don't want to get involved in your argument, but when you say that your images are the same pixel size whether you shoot 12 or 48MP and say that you understand the settings, it should be obvious that something isn't right.

 

[Wreckhunter]

I don't want to get involved in your argument, but when you say that your images are the same pixel size whether you shoot 12 or 48MP and say that you understand the settings, it should be obvious that something isn't right.

And all I did was ask for any ideas,and what do I get.
And there is no argument here.
My opinion and I am not alone here is that the 48mp setting
is not that great.
There sure are some dandies in this forum.
Sorry Im not perfect.

 

[brett8883]

I went to the Lake two days ago with my air 2 and shot stills in 12 and the 48 mp setting and found no difference at all.
This was just my latest test with the 48 mp setting.
They should have left the camera at just 12 mp which is what the honest mp count is anyway.
I know some argue that this sensor on the air 2 is an honest 48mp sensor.
which it really is not.
My wife processed the photos for me ,as she has a degree in photo work.
She could not tell the difference between the 12 or 48mp setting.
Quad Bayer technology is crap to make a long story short.
Have a good day.

You are comparing a 12MP image from a quad bayer sensor to a 48 MP image from a quad bayer sensor. In both cases you are using a quad bayer sensor so it really doesn’t make logical sense to say the quad bayer isn’t any good based on that.

Sensor size plays a fundamentally important role in image quality for multiple reasons and with the MA2 we are talking cell phone camera sized sensor here. In fact some camera sensors are much larger. That should be the standard of expectation for the MA2 and which it should be compared with. Cell phones… By that metric, from what I have seen, it far exceeds that benchmark and I challenge anyone to argue otherwise.