Hey comunity!

Did you noticed as I've noticed that 4K video in Osmo Action is just upscaled 2.7k with added digital sharpness?
I took video samples and compared them side by side with 200% zoom. 4K is a fake. And I know what I'm saying as I'm 3D Artist and graphics designer at games industry, I see these things. However, you don't need to be a professional to see how similar both dji resolutions are. And no matter if it's stabilized or not.

Just take a video of let's say tree branches and compare them with 200% zoom.
4k has more contrast with thin white outline around (digital sharpness) but with the same amount of detail as 2.7k video.

This would also make sense why dji didn't allow us to set sharpness manually - people would see how fake 4K is...

None of the action cams are, I just made a video about it yesterday making a new post about it since me and some others here have talked about it.

How should real 4K work with a sensor of 4000x3000 Pixel, with lens correction and Rocksteady activated? That is not possible. I tried to find out what really happens and measured the FOV for 1080p with lens correction and Rocksteady on. Comparing this FOV with the maximal one, I found that around 3000 pixel of 4000 pixel are used. The rest is reserved for the electronic stabilization. For 4K, this 3000 Pixel are upscaled to 3840, for FullHD they are downscaled to 1920.

Helmut Ruch Posted at 11-28 03:38
How should real 4K work with a sensor of 4000x3000 Pixel, with lens correction and Rocksteady activated? That is not possible. I tried to find out what really happens and measured the FOV for 1080p with lens correction and Rocksteady on. Comparing this FOV with the maximal one, I found that around 3000 pixel of 4000 pixel are used. The rest is reserved for the electronic stabilization. For 4K, this 3000 Pixel are upscaled to 3840, for FullHD they are downscaled to 1920.

You also have to put in that original it's 4:3 aspect, but loose even more resolution when cropped to 16:9.

This is why Rocksteady doesn't work with 4:3 aspect ratio.

johansenfoto Posted at 11-28 03:42
You also have to put in that original it's 4:3 aspect, but loose even more resolution when cropped to 16:9.

This is why Rocksteady doesn't work with 4:3 aspect ratio.

When you compare two 16:9 images aspect ratio doesn't really matter. Also 16:9 doesn't mean that you loose details in the middle of image, it's only less veritical FOV.

Overall it makes no sense to me to put upscaled fake 4K video with same amount of detail (or less as the bitrate for both is very same) as 2.7K.
Of course it makes sens if you want to sell it with 4K written on the box, but is that ethical?

flaj Posted at 11-28 03:59
When you compare two 16:9 images aspect ratio doesn't really matter. Also 16:9 doesn't mean that you loose details in the middle of image, it's only less veritical FOV.

Overall it makes no sense to me to put upscaled fake 4K video with same amount of detail (or less as the bitrate for both is very same) as 2.7K.

If you want to learn more about how almost every brands out there do this, read this:
http://elproducente.com/native-r ... ution-how-to-check/

Ok, I get you. But if everybody do this, is it still good practice? It's just lowering the standards.

johansenfoto Posted at 11-28 04:44
If you want to learn more about how almost every brands out there do this, read this:
http://elproducente.com/native-real-vs-fake-4k-resolution-how-to-check/

There was a very (and I mean "very") weird recent development in the Action 2 forums. A DJI tech said Action 2 actually has  4.6K total sensor readout which enables them to get a true 4K crop even with rocksteady(2) on.

Apart from that, yes, it is mathematically impossible to get 4K resolution after %30 crop from a sensor that is already at 4K limit (horizontally at least) at its full readout.

And you are right, Action 1, as well as all 4K Gopros up to 9 have the same issue as far as I know. Not that I condone the practice of course...

flaj Posted at 11-28 03:59
When you compare two 16:9 images aspect ratio doesn't really matter. Also 16:9 doesn't mean that you loose details in the middle of image, it's only less veritical FOV.

Overall it makes no sense to me to put upscaled fake 4K video with same amount of detail (or less as the bitrate for both is very same) as 2.7K.

Sorry, I wanted to reply to your other post It is below, my bad...

flaj Posted at 11-28 05:09
Ok, I get you. But if everybody do this, is it still good practice? It's just lowering the standards.

This has been "out" for quite some time (you can find me complaining about it in so many posts) , and it seems like people don't care.

Sad, but true...

well  lie about it, get burnt, and DJI  Staff here won't reply to it they rather stay out ofit.  Ever since the 90's three cameras came out they had Interpolated software enhanced, and still do.   I doubt these 23Mp action camers are more than 5Mp

johansenfoto Posted at 11-28 04:44
If you want to learn more about how almost every brands out there do this, read this:
http://elproducente.com/native-real-vs-fake-4k-resolution-how-to-check/

Good read.  Thanks for sharing.

flaj Posted at 11-28 05:09
Ok, I get you. But if everybody do this, is it still good practice? It's just lowering the standards.

Not a good practice and definitely not ethical.  

It is exactly why I'm shooting 2.7K
And to be totally honest, 95% of the views won't need 4K.

Montfrooij Posted at 11-28 09:41
It is exactly why I'm shooting 2.7K
And to be totally honest, 95% of the views won't need 4K.

That is true, it's not easy to see the difference betweek 2K and 4K on the small sensors nowdays.

Like my newest snowplowing video, I had to put the drone to 1440p to get 60 fps, and it still look really good.

johansenfoto Posted at 11-28 12:51
That is true, it's not easy to see the difference betweek 2K and 4K on the small sensors nowdays.

Like my newest snowplowing video, I had to put the drone to 1440p to get 60 fps, and it still look really good.

So far I did not have any real use for 4K.
Nor did any of my camera's

Montfrooij Posted at 11-28 12:59
So far I did not have any real use for 4K.
Nor did any of my camera's

I want the best quality as possible

Btw I got Gopro 10 now also, but waiting for warmer weather to test it out.

Fishycomics Posted at 11-28 06:22
well  lie about it, get burnt, and DJI  Staff here won't reply to it they rather stay out ofit.  Ever since the 90's three cameras came out they had Interpolated software enhanced, and still do.   I doubt these 23Mp action camers are more than 5Mp

Unfortunately most of spec lies are "safe" ones, so you cannot take action (no pun intended) against the companies.

For instance they put a real 20+ megapixel sensor in a camera despite knowing that it would never reach that kind of definition with the given lens setup (due to diffraction limit). It does not work as intended, but it is there.

Quad Bayer sensors are another arguable batch, which actually have high megapixel sensors, anywhere between 48 to 108, but the color resolution is only 1/4th (1/9th in case of 108MP) of a standard sensor. Since color is always interpolated (albeit at a reasonable level) they can get away with whatever spec naming they like. Legally, the base sensor resolution is there. Even the term "Quad Bayer" is a disgrace as the Bayer filters in these sensors are actually "quarter" of a regular one. They can call it "Quad", which sounds higher, better, because there is 1 color filter for every 4 pixels.

So even if the final output is nothing like what it should be, since the advertised hardware exists in some form or another, you cannot do anything apart from shaming and boycotting these products...

Interpolated 4K is a different evil though... By the way if my memory serves me right, until Action 1 came around Gopro's maximum stabilized output was 2.7K, as that was about what was left after the crop. Starting with Hero 9 they don't have the this problem, at least on paper.

For osmo action with a 4000x3000 sensor, if u shooting with rocksteady off. you will get a 4K video output. No crop no Upsampling.

For any 4K video,  no matter use a DSLR or any camera.  Any kind of EIS in post production requires a crop , since it needs space to do the steady.
For a realtime EIS, the difference is no need for post production,  a steady video is generated the moment when finishing recording.  But method is the same.
The other way is to use mechanial gimbal for a 4K DSLR to get a 4K steady video with no crop, which is heavy , inconvenient.

For action2 , we use 4.6K sensor ouput to generate a steady 4K video, which contributes the rise the image quaility .
But for a camera sensor, same sensor area for more resolution  or bigger pixel size ? it is a trade off thing.

P.S. The sharpness thing. you can always use cinelike to shoot a video with little sharpness for post processing.

DJI-ytao Posted at 11-28 18:51
For osmo action with a 4000x3000 sensor, if u shooting with rocksteady off. you will get a 4K video output. No crop no Upsampling.

For any 4K video,  no matter use a DSLR or any camera.  Any kind of EIS in post production requires a crop , since it needs space to do the steady.

Thanks a lot for the conformation DJI-ytao; and as I said before, really nice to hear that you went with a 4.6K sensor instead in Action 2.

Our main problem here is not the technical limitations (which obviously applies to all brands), but rather the presentation of the stabilized 4K video in Action 1. All our estimates (which you can find in multiple threads) bring the hypersmooth crop in Action 1 to something much closer to 2.7K then 4K. Please correct if we are wrong of course. And if we are not, then it should have been presented as a 2.7K output, instead of an interpolated 4K; that is where we have the problem.

I personally think below 1", the smallest resolution possible that leaves a solid 4K crop after stabilization is the best solution, so kudos to Action 2 for that

johansenfoto Posted at 11-28 13:33
I want the best quality as possible

Btw I got Gopro 10 now also, but waiting for warmer weather to test it out.

I don't believe 4K is better quality.
It is more pixels.
But most of the time, it is just more blurred pictures. Nothing more.

DJI-ytao Posted at 11-28 18:51
For osmo action with a 4000x3000 sensor, if u shooting with rocksteady off. you will get a 4K video output. No crop no Upsampling.

For any 4K video,  no matter use a DSLR or any camera.  Any kind of EIS in post production requires a crop , since it needs space to do the steady.

As I mentioned in 1st post I was comparing also non stabilised video and my conclusions are the same. There are very slightly differences with basicly upscaled image and added sharpness.

Montfrooij Posted at 11-28 23:53
I don't believe 4K is better quality.
It is more pixels.
But most of the time, it is just more blurred pictures. Nothing more.

I know but..

johansenfoto Posted at 11-29 01:55
I know but..

Best to give it some tests and don't believe any numbers.

so now do we say the Action 2 is Fake too.  not disagreeing or agreeing.

Fishycomics Posted at 11-29 03:51
so now do we say the Action 2 is Fake too.  not disagreeing or agreeing.

According to post #19 Action 2 has enough resolution to get a full 4K crop after stabilization. So at least on paper it can have true 4K, unlike Action 1. That being said, Action 2 seems to have worse distortion than Action 1, so the correction (dewarp) probably takes a bigger toll...

So should i use 2.7 or 4k to get best video quality with rocksteady? and goes it better if i downscale the footage to 1080p?

dylo Posted at 2-8 16:15
So should i use 2.7 or 4k to get best video quality with rocksteady? and goes it better if i downscale the footage to 1080p?

I still shoot and export in 4K, mostly because I often combine footage from other cameras aswell in my videos.