Does Pocket 2 has an optical zoom capability?

Or is it a digital zoom using 64MP sensor?

Also if it 2X zooms in 4K video, the video quality is lowered? (Just cropping in the 4K video?)

Or it actually utilize 64MP sensor and its high resolution to only use centor 12MP sensor portion to capture actual pixel matching 4K video?



(If it is just digital cropping to half the resolution, why use 64MP instead of 1/1.7" 12MP sensor which would have better low light performance.)

I’ve seen one video today claiming this is lossless zoom, but my guess is similar to Mavic Air2, and although it will suffice and if not zoomed in to much is usable but has a definite loss.

I think it’s the sensor and 12mp used the 64mp is just a Sony quad bayer sensor and is only used for taking photos, it’s basically breaking down the pixels in 4s so 4x16 gives 64 mp but in reality they are not mega pixels but broken into 4 . It’s a bit of trickery, but having used it on Mavic Air 2 , in really good light photographs can show some notable improvements, but good light is important. There is a lot of stuff on google about Sony quad bayer sensors .

hallmark007 Posted at 10-20 12:01
I’ve seen one video today claiming this is lossless zoom, but my guess is similar to Mavic Air2, and although it will suffice and if not zoomed in to much is usable but has a definite loss.

I think it’s the sensor and 12mp used the 64mp is just a Sony quad bayer sensor and is only used for taking photos, it’s basically breaking down the pixels in 4s so 4x16 gives 64 mp but in reality they are not mega pixels but broken into 4 . It’s a bit of trickery, but having used it on Mavic Air 2 , in really good light photographs can show some notable improvements, but good light is important. There is a lot of stuff on google about Sony quad bayer sensors .

Yeah, IMX686 spec chart shows 4 pixel binning.
Then I can understand why 4K only 2X zoom because 16MP is only enough for 2X 4K (8.3MP).

I also heard IMX686 shoots 64MP video in good light condition and shoots 16MP in low light condotion. But not quite sure.

The sensor is a 64MP sensor with the color resolution of a 16MP one. "Quad Bayer" is a controversial marketing name for "quarter resolution" bayer (color) filter. They can use that deceptive name because each filter element is 4 pixels big instead of 1.

With the sensors of this type (which includes the 1/2" 48MP Mavic Air 2 one) individual photosites are unusably small under most light conditions. So the sensor generally bins every 2x2 pixel area to imitate a bigger photosite sensor; bringing the effective megapixel count down to 16MP in DP2's case (and 12MP in MA2's).  That also happens to be the resolution of the bayer filter, so all is well for DP2 at 16MP

If all the pixels are used individually, like when taking a full resolution photo, the colors are highly interpolated.

CemAygun Posted at 10-20 17:03
The sensor is a 64MP sensor with the color resolution of a 16MP one. "Quad Bayer" is a controversial marketing name for "quarter resolution" bayer (color) filter. They can use that deceptive name because each filter element is 4 pixels big instead of 1.

With the sensors of this type (which includes the 1/2" 48MP Mavic Air 2 one) individual photosites are unusably small under most light conditions. So the sensor generally bins every 2x2 pixel area to imitate a bigger photosite sensor; bringing the effective megapixel count down to 16MP in DP2's case (and 12MP in MA2's).  That also happens to be the resolution of the bayer filter, so all is well for DP2 at 16MP  

Thanks for in depth information.
Then does 2x2 pixel binning has any advantage in reducing ISO noise over one big photo-site?

ISO noise happens when each photo-site lacks proper illumination and predicting the result with error. If 2x2 binning uses 3 photo-site for luminosity, averaging out 3 data helps in any way?

Simply thinking, 2x2 binning would have loss of detecting area because of the seam between the photo-site that wouldn’t be there if it was one big pixel  

Tide Posted at 10-20 18:11
Thanks for in depth information.
Then does 2x2 pixel binning has any advantage in reducing ISO noise over one big photo-site?

You are welcome My money would be on that binned ones cannot match a proper big photosite.  But there are so many processing tricks they employ with the pixels that share the same color square, it is hard to tell for sure.  In the end this is also how they get the HDR video.

They list the photosite size as 8μm for individual pixel and 1.6μm binned, as if there is no loss.  So my guess is that the inter-site gap might be negligible.

Above all there is the diffraction limit problem, which no  software trick can cure (Highly oversimplified: the resolution of the light hitting the sensor can be lower than your "pixel count" causing a true output resolution lower than what's on paper). I believe DP2 would be suffering from that at 64MP (when binning is not used) but how big is the loss, I am not sure.

Hello Tide, thank you for reaching out and for your inquiry. The Osmo Pocket 2 has digital zoom using 64MP sensor and zooming can be achieved by cropping the resolution. Thank you.

DJI Paladin Posted at 10-21 07:44
Hello Tide, thank you for reaching out and for your inquiry. The Osmo Pocket 2 has digital zoom using 64MP sensor and zooming can be achieved by cropping the resolution. Thank you.

Thanks DJI Paladin for reply.
But I heard some Youtube reviwers say DJI documents they got with review unit mention 2x optical zoom and 4x digital zoom.

If it is just crop 4K video should be 4x zoom and 1080P video 2x because 4K has more resolution to crop. Why the other way around?

Tide Posted at 10-22 19:51
Thanks DJI Paladin for reply.
But I heard some Youtube reviwers say DJI documents they got with review unit mention 2x optical zoom and 4x digital zoom.

The crop mentioned here is the crop from the full sensor readout into the desired resolution. So if you are recording 4K for instance: Your full horizontal sensor readout is 9216 pixels wide; so you can crop (zoom) into that 2.4 times until you reach 3840 pixels, which is the width of 4K video. Anything more and you would not be recording "true" 4K anymore.

If you do the same math, you can zoom much more into the same sensor readout when you are recording 1080p, as all you need left is 1080 pixels across out of 9216. Vertical resolution is not significant as the sensor is much bigger vertically than the the standard 16:9 video ratio sor there is significant extra margin there.

Of course this is in an ideal situation where the sensor/lense combo can actually resolve 64 mega pixels perfectly, which is not necessarily the case with Pocket 2. But the same principles apply.

Hope this helps

Cheers

CemAygun Posted at 10-23 13:13
The crop mentioned here is the crop from the full sensor readout into the desired resolution. So if you are recording 4K for instance: Your full horizontal sensor readout is 9216 pixels wide; so you can crop (zoom) into that 2.4 times until you reach 3840 pixels, which is the width of 4K video. Anything more and you would not be recording "true" 4K anymore.

If you do the same math, you can zoom much more into the same sensor readout when you are recording 1080p, as all you need left is 1080 pixels across out of 9216. Vertical resolution is not significant as the sensor is much bigger vertically than the the standard 16:9 video ratio sor there is significant extra margin there.

Yeah Sensor crop is what I figured.
Then we should see how much of brightness is affected by zoom.

Tide Posted at 10-23 18:51
Yeah Sensor crop is what I figured.
Then we should see how much of brightness is affected by zoom.

I am very curious myself. Logic states the difference between the binned output and the full 64MP readout should be significant, but who knows...