Google phone versus $3500 Sony FF
 

On 24/11/2018 20:55, RichA wrote:
https://petapixel.com/2018/11/24/goo...ight-shootout/

I have no idea how these really do, but look at the shot of the scene
with water and clouds. Notice the motion blur with the Sony versus the
phone?
==========================================

I have one of the Pixel 3 phones, and confirm how good the low light
capability actually is. Indeed, it was the camera features which made
me choose that phone over any other - it's about three times the price
of the last phone I bought.

Effectively it is just long exposure, but done in a way that camera
shake is less important than it might otherwise be. I have no
expectation that it's anything like as good as even my MFT camera with
the f/1.7 pancake lens, but I have it woth me 24 hours a day, and that's
important!

Folks may be interested in:


https://petapixel.com/2018/11/14/goo...near-darkness/

and the details from Google:


https://ai.googleblog.com/2018/11/ni...-on-pixel.html

No reason why some of these features might start appearing in other
cameras, of course (except patents, perhaps?).

--
Cheers,
David
Web: http://www.satsignal.eu

Google phone versus $3500 Sony FF
 

In article , David Taylor
wrote:


I have one of the Pixel 3 phones, and confirm how good the low light
capability actually is. Indeed, it was the camera features which made
me choose that phone over any other - it's about three times the price
of the last phone I bought.

Effectively it is just long exposure, but done in a way that camera
shake is less important than it might otherwise be.

it's not a long exposure. it's stacked multiple exposures where the
shutter speed is dynamically optimized based on subject movement, which
are then combined using machine learning.

I have no
expectation that it's anything like as good as even my MFT camera with
the f/1.7 pancake lens, but I have it woth me 24 hours a day, and that's
important!

it's likely better.

Google phone versus $3500 Sony FF
 

In article ,
RichA wrote:

Since the m4/3rds camera (thanks to great IBIS) can shoot about a second long
exposure at f/1.7, likely it will take a better night shot than the phone.

not likely, because the phone is using computational photography and
the m4/3rds camera is stuck in the past.

Dpreview compared a night-mode video camera with a Nikon D5. Problem was,
the D5 and the lens they used had no stabilization so it was limited to
1/30th without blurring the image, at 20,000 ISO. Dual IBIS and a 50mm lens
will yield practical exposures of seconds long so unless there is fast
movement in the image, it's a better solution. Not only that, you can drop
the ISO so the image doesn't look like crap.

not relevant.

Google phone versus $3500 Sony FF
 

On 26/11/2018 10:01, RichA wrote:
[]
Since the m4/3rds camera (thanks to great IBIS) can shoot about a second long exposure at f/1.7, likely it will take a better night shot than the phone.

Yes, that's my expectation as well, as I said.

DPreview compared a night-mode video camera with a Nikon D5. Problem
was, the D5 and the lens they used had no stabilization so it was
limited to 1/30th without blurring the image, at 20,000 ISO. Dual IBIS
and a 50mm lens will yield practical exposures of seconds long so unless
there is fast movement in the image, it's a better solution. Not only
that, you can drop the ISO so the image doesn't look like crap.

As I understand it, the Pixel has, effectively, post-exposure image
stabilisation, having taken a sequence of images. It does this by image
macting rather than movement compensation. That's why I said that other
cameras might be able to use the same technique.

In practice, these days, I don't even have a full-frame DLSR, and I'm
unlikely to carry around my MFT unless I'm on a photograhic expedition.
But I will have my phone, so I'm grateful for the extended facilities it
provides.

--
Cheers,
David
Web: http://www.satsignal.eu

Google phone versus $3500 Sony FF
 

In article , David Taylor
wrote:

As I understand it, the Pixel has, effectively, post-exposure image
stabilisation, having taken a sequence of images. It does this by image
macting rather than movement compensation. That's why I said that other
cameras might be able to use the same technique.

you understand wrong. the pixel uses machine learning to combine
multiple exposures, similar to hdr but optimized for low light. it also
varies the exposure of each photo based on subject movement.

Google phone versus $3500 Sony FF
 

In article ,
RichA wrote:

Since the m4/3rds camera (thanks to great IBIS) can shoot about a second
long
exposure at f/1.7, likely it will take a better night shot than the
phone.

not likely, because the phone is using computational photography and
the m4/3rds camera is stuck in the past.

Computational photography is guesswork, partly, it's like how a CD or DVD
player avoids skipping, except the CD or DVD player can actually read-ahead
whereas the camera cannot predict with precision (most of the time) what a
subject is going to do.

nope.

cd/dvd players use extensive error correction, with cd-rom/dvd-rom even
more so. a minor error in an audio or video stream isn't a big deal,
but *any* error in a data file *is*.

computational photography is not error correction. it computes what
should be there, and the results are quite good and getting better all
the time.

Dpreview compared a night-mode video camera with a Nikon D5. Problem was,
the D5 and the lens they used had no stabilization so it was limited to
1/30th without blurring the image, at 20,000 ISO. Dual IBIS and a 50mm
lens
will yield practical exposures of seconds long so unless there is fast
movement in the image, it's a better solution. Not only that, you can
drop
the ISO so the image doesn't look like crap.

not relevant.

It is if the goal is to produce a better static image such as the one
referenced. No IBIS or stabilization on night shots is crippling, unless you have a tripod.

a nikon d5 and a video camera are not relevant to a phone camera and
isp.