Ill-Advised Reviews: Pixel 3 vs. Latest iPhone XS Zoom
I don't own either of these devices. And I do own a Pixel 2, so I'm probably biased. Hence, "ill-advised".
While I cannot guarantee an absence of bias, I can attest to not exactly being Google's biggest fan. And maybe that offsets things a bit. But, I also do genuinely dislike Apple, so I probably lose any gains automatically.
But here goes anyway.
I've been reading articles about the Pixel 3 because I'm thinking about picking one up when I next get a bit of extra money. Inevitably, we run across articles like this one. What blows my mind is that article effectively distills the differences in these two phones to one aspect of one feature. The zoom on the camera.
Needless to say, there is more to a phone than a camera and more to a camera than zooming.
But, the worst part is that this isn't even a fair comparison. We're talking about 2 different price points. On two different levels of availability on two entirely different technologies.
First up is price. The most affordable Pixel 3 with this feature will be the base Pixel 3 with 64GB storage which goes for $999CAD. The cheapest iPhone XS, $1379CAD. That is more than fully an extra 1/3 the price of the Pixel 3 over again. If this feature is your sole differentiator, then it better be worth an extra $380CAD. And frankly, the differences detailed don't amount to that to me.
So before we dive in further, let's talk about the pictures. We are looking at a super detailed scene with a lot of contrasting colors and a lot of detailed specimen. Oh, and it is in bright, direct sunlight. Basically, the only situation in which iPhone cameras tend to "win" in photo comparisons. I'm still stunned that the article draws it's conclusions so brazenly on a single example.
And what an example; Birds, which, even at 2x zoom, fit what? 50 in a single shot. Highly textured brick. Some cabling. This choice is picture is, simply put, not the typical user photo. And, both cameras do an impressive job. Does the iPhone win? Absolutely. Would any average phone user be disappointed with either? I think not. But, not really the point. We've yet to see how the two zoom features (let alone the rest of the functionality on the two phones) compare in a broader set of tests.
iPhones tend to suck at low light. Zoom isn't going to make that better. And even in decent lighting situations where other phones tend to beat iPhones, how will the two compare? Simply put; we don't know.
But, as stated, the picture is a good technical test scene in environments which have always tended to favor Apple's image processing and default camera settings. But, it doesn't even end there.
Next is availability. The XS is available. The Pixel 3 is not publicly available yet. And these tests were not necessarily even done with final version of the software, whereas iPhones have not only been released but also updated since launch. While I doubt that will change much in the span of a few weeks, you can't really go out and make comparisons like this between a production unit and an unreleased one.
The final point is near and dear to me. When I heard how the Pixel 3 managed it's digital zoom, I was blown away. I had similar thoughts in the past and wondered why no one had done something like this yet. And maybe the software and hardware just weren't there yet. But, it is inescapable; the iPhone isn't using digital zoom. It is using optical zoom. The quality of the zoom will simply be dependent on the quality of the lens and the camera. And yeah, Apple uses high quality products. As stated above though, you're paying for it. And, as also stated above, if there are conditions where the lens or camera don't perform, the zoom will actually amplify the deficiencies, which is why I'm REALLY say to see a single shot which BLATANTLY panders to Apple's strengths and then draws a conclusion.
Google's solution is brilliant. But it is software. And software like this isn't perfect. It is just REALLY good.
I'm also interested in the testing methodology. For instance, was a tripod used? If so, it may not be generating enough variation in the multiple shots to scrub for additional details.
Anyway, end of the day, optical zoom doesn't require processing, cropping or tricks to accomplish the zoom. It will always be the most able to provide the truest representation. Digital zooms on the other hand rely on sacrifices or trickery. Without optical zoom, you're always trying to blow a smaller piece of an image up with any additional data to improve the quality.
Well, sort of at least. Google actually IS using additional data. It just isn't at any better resolution or zoom than the original.
If this feature alone were enough to make or break a phone (and it isn't). Apple should be shitting their pants. Because if this were also applied to a device with say, optical zoom and multiple lens... it would tear Apple a new one. What Google achieved here is insane. For digital zoom to look so good is downright incredible.
For those who don't know. What Google is basically doing is buffering a bunch of images, then, when you take a picture, it grabs some number of pictures from before and after, and uses the differences caused in each shot by your natural inability for your wimpy human hands to hold a camera perfectly still, and uses the extra data, when it is clean enough, to help it figure out and fill in more details. Those little variations in movements may put a bit more into focus in a pixel in one shot which wasn't there in another. And group of pixels may be clearer in one shot than another. Data like this can also be used to help determine depth with just a single lens which can really help improve algorithms. All of these deltas in the data between the collection of images allow the algorithm to basically create an image with a virtually inflated megapixel count.
And, there is definitely going to be room for improvement in these algorithms.
Here is the craziest thing about this technology. While optical zoom doesn't have to worry about algorithms. It is still limited in many ways. The sum total raw image data the Pixel 3 gets may not be guaranteed to be "true to life". But, it can (theoretically) produce a more accurate picture in the right conditions.
To illustrate, let's get insanely basic. Imagine 2 different 1 pixel cameras, one with digital zoom and one with optical zoom. Not 1 megapixel. Just ONE pixel. A one Pixel camera will generally produce a picture with a single pixel. A single point of color. If you shoot a picture of the ocean on these cameras from far away, that pixel will probably end up roughly the median color of the ocean from where you are. The camera with optical zoom though, might be able to zoom in really close and get a white pixel from a far out white cap. Whereas the digital zoom camera will always return the same color. It can't zoom to get any more details.
What the Pixel 3 does, is it takes 15 pictures. So, it would actually have 15 pixels worth of data. If each of those pixels just so happened to be adjacent, it could actually produce a 15 pixel picture. Or put another way, an image with a 1500% greater resolution than the 1 pixel camera with optical resolution. But, what is more likely to happen than getting 15 pixels all uniformly distributed, is you'll get a bunch of them overlapping with every so slightly different shades more times. And you'll probably end up with say, a 5 pixel shot with each pixel being the average data from 3 overlapping pixels. And, which any luck you'll not only get a greater resolution photo, but also, potentially a more accurate photo.
The Pixel 3 camera in Super Res Zoom is doing something like this, but across all 12 million pixels. If the data is well varied and the algorithms are tuned appropriately this is how it can actually beat out a superior physical camera. Of course, if the data is garbage and/or the algorithms suck or hit a snag on a certain data set, it could ruin an otherwise good photo.
I imagine that this close to launch, the Pixel 3 hardware and algorithms are closer to the good end. But, not as good as they can get. Which is why I'm a bit peeved about the comparison to a phone which isn't officially available, in conditions which favor the incumbent historically and where the incumbent is also a more mature product (in that it has been live in the market longer and had longer to receive updates).
I don't expect the Pixel 3 to EVER beat the iPhone XS is lighting and conditions like the ones in the linked article. With the ideal lighting conditions (bright natural sunlight, behind the camera) it will be a long time before software can bridge the gap to reliably beat out optical zoom. I FULLY expect that in worse conditions, the XS will actually get spanked, even at zoom.
As with the single pixel camera example. The iPhone still only has one picture of one pixel. If it is the wrong picture, nothing can save it. If the Pixel 3 takes 15 shots and they are all total crap except 1, and the algorithms detect that, it can still just serve up that 1 photo untouched. But, more than likely it will have more than one good frame to work with, and thus should generally be able to improve the image quality in more difficult scenarios.
While I cannot guarantee an absence of bias, I can attest to not exactly being Google's biggest fan. And maybe that offsets things a bit. But, I also do genuinely dislike Apple, so I probably lose any gains automatically.
But here goes anyway.
I've been reading articles about the Pixel 3 because I'm thinking about picking one up when I next get a bit of extra money. Inevitably, we run across articles like this one. What blows my mind is that article effectively distills the differences in these two phones to one aspect of one feature. The zoom on the camera.
Needless to say, there is more to a phone than a camera and more to a camera than zooming.
But, the worst part is that this isn't even a fair comparison. We're talking about 2 different price points. On two different levels of availability on two entirely different technologies.
First up is price. The most affordable Pixel 3 with this feature will be the base Pixel 3 with 64GB storage which goes for $999CAD. The cheapest iPhone XS, $1379CAD. That is more than fully an extra 1/3 the price of the Pixel 3 over again. If this feature is your sole differentiator, then it better be worth an extra $380CAD. And frankly, the differences detailed don't amount to that to me.
So before we dive in further, let's talk about the pictures. We are looking at a super detailed scene with a lot of contrasting colors and a lot of detailed specimen. Oh, and it is in bright, direct sunlight. Basically, the only situation in which iPhone cameras tend to "win" in photo comparisons. I'm still stunned that the article draws it's conclusions so brazenly on a single example.
And what an example; Birds, which, even at 2x zoom, fit what? 50 in a single shot. Highly textured brick. Some cabling. This choice is picture is, simply put, not the typical user photo. And, both cameras do an impressive job. Does the iPhone win? Absolutely. Would any average phone user be disappointed with either? I think not. But, not really the point. We've yet to see how the two zoom features (let alone the rest of the functionality on the two phones) compare in a broader set of tests.
iPhones tend to suck at low light. Zoom isn't going to make that better. And even in decent lighting situations where other phones tend to beat iPhones, how will the two compare? Simply put; we don't know.
But, as stated, the picture is a good technical test scene in environments which have always tended to favor Apple's image processing and default camera settings. But, it doesn't even end there.
Next is availability. The XS is available. The Pixel 3 is not publicly available yet. And these tests were not necessarily even done with final version of the software, whereas iPhones have not only been released but also updated since launch. While I doubt that will change much in the span of a few weeks, you can't really go out and make comparisons like this between a production unit and an unreleased one.
The final point is near and dear to me. When I heard how the Pixel 3 managed it's digital zoom, I was blown away. I had similar thoughts in the past and wondered why no one had done something like this yet. And maybe the software and hardware just weren't there yet. But, it is inescapable; the iPhone isn't using digital zoom. It is using optical zoom. The quality of the zoom will simply be dependent on the quality of the lens and the camera. And yeah, Apple uses high quality products. As stated above though, you're paying for it. And, as also stated above, if there are conditions where the lens or camera don't perform, the zoom will actually amplify the deficiencies, which is why I'm REALLY say to see a single shot which BLATANTLY panders to Apple's strengths and then draws a conclusion.
Google's solution is brilliant. But it is software. And software like this isn't perfect. It is just REALLY good.
I'm also interested in the testing methodology. For instance, was a tripod used? If so, it may not be generating enough variation in the multiple shots to scrub for additional details.
Anyway, end of the day, optical zoom doesn't require processing, cropping or tricks to accomplish the zoom. It will always be the most able to provide the truest representation. Digital zooms on the other hand rely on sacrifices or trickery. Without optical zoom, you're always trying to blow a smaller piece of an image up with any additional data to improve the quality.
Well, sort of at least. Google actually IS using additional data. It just isn't at any better resolution or zoom than the original.
If this feature alone were enough to make or break a phone (and it isn't). Apple should be shitting their pants. Because if this were also applied to a device with say, optical zoom and multiple lens... it would tear Apple a new one. What Google achieved here is insane. For digital zoom to look so good is downright incredible.
For those who don't know. What Google is basically doing is buffering a bunch of images, then, when you take a picture, it grabs some number of pictures from before and after, and uses the differences caused in each shot by your natural inability for your wimpy human hands to hold a camera perfectly still, and uses the extra data, when it is clean enough, to help it figure out and fill in more details. Those little variations in movements may put a bit more into focus in a pixel in one shot which wasn't there in another. And group of pixels may be clearer in one shot than another. Data like this can also be used to help determine depth with just a single lens which can really help improve algorithms. All of these deltas in the data between the collection of images allow the algorithm to basically create an image with a virtually inflated megapixel count.
And, there is definitely going to be room for improvement in these algorithms.
Here is the craziest thing about this technology. While optical zoom doesn't have to worry about algorithms. It is still limited in many ways. The sum total raw image data the Pixel 3 gets may not be guaranteed to be "true to life". But, it can (theoretically) produce a more accurate picture in the right conditions.
To illustrate, let's get insanely basic. Imagine 2 different 1 pixel cameras, one with digital zoom and one with optical zoom. Not 1 megapixel. Just ONE pixel. A one Pixel camera will generally produce a picture with a single pixel. A single point of color. If you shoot a picture of the ocean on these cameras from far away, that pixel will probably end up roughly the median color of the ocean from where you are. The camera with optical zoom though, might be able to zoom in really close and get a white pixel from a far out white cap. Whereas the digital zoom camera will always return the same color. It can't zoom to get any more details.
What the Pixel 3 does, is it takes 15 pictures. So, it would actually have 15 pixels worth of data. If each of those pixels just so happened to be adjacent, it could actually produce a 15 pixel picture. Or put another way, an image with a 1500% greater resolution than the 1 pixel camera with optical resolution. But, what is more likely to happen than getting 15 pixels all uniformly distributed, is you'll get a bunch of them overlapping with every so slightly different shades more times. And you'll probably end up with say, a 5 pixel shot with each pixel being the average data from 3 overlapping pixels. And, which any luck you'll not only get a greater resolution photo, but also, potentially a more accurate photo.
The Pixel 3 camera in Super Res Zoom is doing something like this, but across all 12 million pixels. If the data is well varied and the algorithms are tuned appropriately this is how it can actually beat out a superior physical camera. Of course, if the data is garbage and/or the algorithms suck or hit a snag on a certain data set, it could ruin an otherwise good photo.
I imagine that this close to launch, the Pixel 3 hardware and algorithms are closer to the good end. But, not as good as they can get. Which is why I'm a bit peeved about the comparison to a phone which isn't officially available, in conditions which favor the incumbent historically and where the incumbent is also a more mature product (in that it has been live in the market longer and had longer to receive updates).
I don't expect the Pixel 3 to EVER beat the iPhone XS is lighting and conditions like the ones in the linked article. With the ideal lighting conditions (bright natural sunlight, behind the camera) it will be a long time before software can bridge the gap to reliably beat out optical zoom. I FULLY expect that in worse conditions, the XS will actually get spanked, even at zoom.
As with the single pixel camera example. The iPhone still only has one picture of one pixel. If it is the wrong picture, nothing can save it. If the Pixel 3 takes 15 shots and they are all total crap except 1, and the algorithms detect that, it can still just serve up that 1 photo untouched. But, more than likely it will have more than one good frame to work with, and thus should generally be able to improve the image quality in more difficult scenarios.
Comments
Post a Comment