[dgacmu]

I'm sorry, but while you may be a photography fan, you don't know what you're talking about here.

Stacking is quite the opposite of a "ghetto" version of a long exposure - it's the fundamental building block of being able to do the equivalent of a long exposure without its associated problems (motion blur from both camera and subject, high sensor noise if you turn up the gain, and over-saturating any bright spots).

Stacking is the de facto technique used for DSLR astrophotography for exactly these reasons -- see https://photographingspace.com/stacking-vs-single/

However, you're ignoring the _very substantial_ challenges of merging many exposures taken on a handheld camera. Image stabilization is great, but there's a lot of motion over, say, 1 second on a hand-held camera. Much more than the typical IS algorithm is designed to handle.

The techniques are non-trivial: http://graphics.stanford.edu/talks/seeinthedark-public-15sep...

There's a lot going on to accomplish this. It starts with the ability to do high-speed burst reads of raw data from the CCD (so that individual frames don't get motion blurred, and raw so you can process before you lose any fidelity by RGB conversion), and requires a lot of computational horsepower to perform alignment and do merging. I don't know what the Pixel's algorithms are, but merging of many images with hand-held camera motion benefits from state of the art results in applying CNNs to the problem, at least, from some of the results from Vladlen Koltun's group at Intel (who I'd put at the forefront of this, along with Marc Levoy's group at Google):

http://vladlen.info/publications/learning-see-dark/

I wouldn't be so quick to dismiss the technical meat behind state of the art low-light photography on cell phones.

[endorphone]

"you don't know what you're talking about here"

You literally repeated exactly what I said image stacking was, yet lead off by claiming that I don't know what I'm talking about. Classic.

The goal of both is to achieve the exact same result -- more photons for a given pixel. Stacking is a necessary compromise under certain circumstances -- lack of sufficient stabilization, particularly noisy sensor or environment, etc.

Further, this implementation is clearly long exposures (note the blur rather than strobe).

[dgacmu]

Sorry, but no - stacking is as much about dynamic range adaptation as noise, and that's why I'm arguing against your terming it a "ghetto version of a long exposure". It's not. A long exposure has a fundamental problem with saturation, as well as noise. It's not just about lack of stabilization, there's also the motion of the subject. Computational approaches can compensate for subject motion - long exposures cannot. Computational approaches can do dynamic range adaptation to avoid blow-out due to CCD pixel saturation. Long exposures cannot.

If you read the slides from the Levoy talk I cited, you'll note that they explicitly choose to under-expose the individual exposures to minimize motion blur and blowout.

(Marc is now at Google continuing his work on computational photography, and his group contributes to many of the cool things you see on the Pixel series.)

[endorphone]

"Computational approaches can compensate for subject motion"

But they don't. They don't in this example. Moving subjects are a blur. As an aside, of course stacked photo frames are underexposed because it wouldn't make much sense otherwise.

Computational photography can do interesting things and holds a tremendous amount of promise. However every single example that I can find of this mode -- across the many astroturfed pages -- show a longer exposure than what the stock app normally allows. And with that the requisite blurring of any moving subject.

[hug]

Are you arguing that EIS over a series of burst photos is incapable of making things better?

Everything I can see you saying -- much if it agreeable, like the fact that long-exposure OIS makes a lot of what this technology currently does possible without it -- is simply handwaving away the fact that EIS-over-burst with OIS can achieve things that OIS cannot by itself.

It seems to me that it's patently true that EIS has some benefits, and those benefits can be realised over the top of OIS.

There's obviously still a fair limit to OIS. I have somewhat shaky hands and even using something like Olympus' top range 5-axis IBIS, which is the best I've ever seen, I can still only shoot at 1/10". What can EIS do with a burst of 3x 1/20" exposures? Probably counter for my shaking a bit, at least. (If not for subject movement, yet.)

I simply do not see why you're discounting this so heavily.

[endorphone]

Where did I discount EIS? EIS+OIS is a golden solution. It's what the Pixel 3 does. It's what the iPhone 8 does. It's what the Huawei P20 does.

This all gets very reductionist, but EIS over a series of bursts is a bad alternative to OIS. It will be garbage in->garbage out. EIS with OIS, however, gives you the benefits of OIS, with the safety valve and "time travelling" effect of EIS (in that it can correct where OIS made the wrong presumption, like the beginning of a pan).

>and even using something like Olympus' top range 5-axis IBIS

The ability of OIS to counter movement is a function of the focal length. Your Olympus probably has a 75mm equivalent or higher lens, where a small degree of movement is a large subject skew. That smartphone probably has a 26-28mm equivalent lens. Small degrees of movement are much more correctable.

EIS is brilliant. OIS is better for small movements, but add EIS and it's great. Computational photography is brilliant. However Google has really, really been pouring out the snake oil for their Pixel line.

[hug]

> Your Olympus probably has a 75mm equivalent or higher lens

50mm equivalent in 135 terms, but yes, larger than 28. (I've since moved on to an X100F, but that's neither here nor there. :)

> EIS is brilliant. OIS is better for small movements, but add EIS and it's great. Computational photography is brilliant. However Google has really, really been pouring out the snake oil for their Pixel line.

This is what I was missing. It seemed you were arguing that computational photography is not capable of much, but you're more just pointing out that this computational photography is not doing much, despite Google's claims to the contrary.

I'd agree with you that this is not exactly revolutionary stuff.

[ricardobeat]

> But they don't. They don't in this example. Moving subjects are a blur.

If you stack the original exposures together, you'll get ghosting and not a blur. The natural-looking blur is a result of computation.

> of course stacked photo frames are underexposed, wouldn't make sense otherwise

Except it does make sense if you want to capture more shadow detail, this is how HDR images are made.

You're severely underestimating the amount of computation involved in getting these shots. These are all handheld, and as @dgacmu mentions can benefit from exposure bracketing which gives much better results than a single long exposure.

Of course you could already get similar shots from a good camera and technique - the fact these are handheld shots coming from a mobile device, and straight out with the camera is the impressive part.