31 comments

[ 2.8 ms ] story [ 73.1 ms ] thread
I'm wondering why this is exclusive to Pixel 3 and not made available to Pixel 2 owners? If it's an algorithm that stiches together many images, Pixel 2 should be able to do this as well. The Pixel 2 camera already meshes together many images when you take photos, and already gives users the ability to take "motion" photos.
People have gotten a lot of Pixel 3 exclusive features to work one Pixel 1 + 2 by install Pixel 3's camera apk. I'm assuming most of the reason for backporting features is not a hardware limitation.
Google gets to sell a couple of new Pixel 3 devices?
same reason everyone throws away their phone every year and buys a new one.
Super Zoom needs a better lens than what the sensor strictly requires. That was probably not the case on the Pixel 2.

Also, can the Pixel 2 control the OIS the way the Pixel 3 does? I own both, but have no way of checking.

Specifically if the lens is too good one gets aliasing which cannot be removed through software.

The pixel 3 hardware requires this super zoom method at all times, or the output will be aliased.

Could also be improvements necessary to the "Pixel Visual Core" that are necessary to make this run fast enough.
Google is actually backporting a quite a lot of Pixel 3 functionality to the older phones - e.g. night sight (which makes a enormous difference for photos taken in less than ~3 lux), call screening, adjustable blur. (Full list at https://www.androidpolice.com/2018/10/13/pixel-3-features-co... )

It's pretty generous actually, Apple and Samsung rarely backport marquee features. I figure Google is trying to build up loyalty amongst Pixel and Pixel 2 owners so that they stay with the brand.

> I figure Google is trying to build up loyalty amongst Pixel and Pixel 2 owners so that they stay with the brand.

As a Pixel owner: it's working.

As a nexus/pixel user since the Galaxy Nexus I'd say so too :)
Canon/Nikon should put an Android phone into their cameras. You could get these kind of features/easy instagramming/connection capabilities starting from a much higher photo quality base.

Keep their native interface, but also allow switching to a much slower (in the sense of many touches required) but much more capable Android mode.

I think an actually working quick way of getting photos transferred to a phone out of the box would be the better step. That stuff's optimized for a phone, use a phone and keep it out of the camera, which likely will have a longer lifetime and adds a bunch of other constraints.
But then you wouldn't have features like this one, which requires merging multiple close in time images. Since the professional camera makers assume that all post-processing will happen on a computer at a later time, they don't work on in-camera features like this.
A burst of 150 raw frames all within 1 second even the top end slr's can't do, mostly because they don't have enough RAM to store all those raw frames, and because sd cards aren't fast enough.

Without all those raw images, you won't be able to do this kind of postprocessing after the fact.

You're right that SD cards aren't fast enough. That's why high-end DSLRs support the much faster XQD standard. There are no DSLRs (and precious few if any still cameras, period) that can do 150fps, but 15-20fps isn't at all uncommon in professional lines, and most cameras that support that frame rate also have deep enough buffers and fast enough XQD interfaces that they can shoot raw at full framerate for 10-15 seconds before they fill the buffer and slow down.
In-camera exposure stacking has been a standard feature in DSLR firmware for some years now.
Too bad this is from Google, so I can't use this; neither at work nor at home.
Do you live in a totalitarian state?
I think that's what the OP is trying to avoid.
Google online products are pretty much blocked in China.

This is on-device software though, so probably not an issue.

Do you work at Apple/Amazon/Facebook?
I find it interesting that this tech is only just becoming available for flagship phones. It seems to be the same algorithm that BlessN900 was using for the Nokia N900, back in 2010: https://web.archive.org/web/20100903131550/http://blessn900....
In retrospect, the N900 was such a visionary product. Maybe not as much as the Palm Pre, but close to it IMO.
In what way other than the camera? Nokia flagships like this, the N95, N8 and more in that generation had some excellent cameras. The N900 had a resistive touchscreen at the same time the iPhone 3GS and Nexus One were available. The OS was a blessing and a curse.
One nice thing about the N900's resistive touchscreen is that it was pressure-sensitive. This was great for drawing apps (notably MyPaint), and I haven't really seen a similar drawing experience on a phone apart from Apple's Force Touch.
True. Ability to use with gloves as well. Apple could probably make the Apple Pencil work with iPhones, but it probably wouldn’t be worth the hassle.
Whilst this is genuinely amazing work and very exciting, I would just like to point out a few things.

The comparable tech in digital cameras is not limited to tripods and static scenes as stated, some cameras can do this hand-held.

Shifting pixels 4 times results in increased detail, however Olympus and Panasonic’s approach is to shift the sensor 8 times in half-pixel increments, resulting in dramatically increased resolution as well.

The main issue with the current implementations in digital cameras is still objects in motion, which camera makers are trying to solve in-software with mixed results.

If rumors are to be believed there are advances on the way for this tech in digital cameras, particularly in relation to shutter speed and perhaps low light, but I don’t think you’ll see this trickle down to cheaper models for a few years yet.